Place your take-home messages here


Part of pay for being here is to show your little brothers and sisters



Download 17.96 Mb.
Page13/57
Date28.02.2021
Size17.96 Mb.
1   ...   9   10   11   12   13   14   15   16   ...   57

Part of pay for being here is to show your little brothers and sisters.

1.3 light seconds.

That is your neighborhood.

Distance to sun is 8 minutes.

Little computer board - Arduino.

I've preprogrammed it, it is talking to you in morse code \"SU\"

Program Arduino to delay signal 2.7 seconds.

Rate of switching is 5 words per minute.

Win this XPrize pin.
Basic idea. Plant a seed of activity in space. Feed it with material from moon.

Build solar power satelites. First make copies of system elements.

Bob Friedas's [ref]

Robot in sea of robot parts making copy of itself.NASA has looked at how you might do this on the moon.

Literally exponential growth.

Take only smallest tool set you require. Make copies of your civilization.

Does not require nano.

Can do it with 60's technology.

What if Earth has abundant energy.

Collectors change albedo of earth.
We travel a lot.

Can you do flying without fuel? a light craft. Lasers around for 50 years.

Plane that heats air.

Fuel in 747 is only used to heat air.

You could heat air with laser.

Illustration of one woman space vehicle powered by laser beam coming from space.

Real copies of that exist. Just not that big [ref]

Lasers are now really cheap.

Solid state lasers obey Moore's Law.

Kare, George - banks of lasers on ground to launch.

Nano is great, but you can do this soon to get into Space.

Solar pumps in space.

List of books:

Gerard O'Neill. Giving this to Jose. [ref]

Wonderful.

Solar sails pushed by light photos.

These sails are pushed by radio waves.
- Approaching the Energy Singularity

Moving one 747 from London to New York uses same energy of 1066 England - total.
Limit will be heat.

If everybody had 10-20x. Be careful about combustion products.

Comes down to policy.

If we don't tax for putting carbon in, it won't be cost effective to do these other things.

We need to make the conscious choice to save our planet.

Smaley.

- Energy Gamechagers

Carbon

Harvested Carbon Fuels

\"New\" nuclear

Cost revolution in renewables

Storage

Ultragrid

Power from Space
- [photo of Earth]

Here is what is at stake. Yours and your kids.
Changed from being a lawyer. Not rearranging marbles in a zero sum game.

The sun is too generous to us.

The game is all about energy. Space has a lot to do with it.

Energy has never been so exciting.

Energy basics later.

12:59 [applause]

Q:Dmitry. Exponential decay of laser price? How compared to microwaves?
Will give you Kare's paper.

Cost now is much lower.
Microwaves - radio energy - extremely efficient 83% at received end. Compare to really good solar cells of 20%. Looks like chain link fence laying on the ground.

Energy density is less than sunlight.

Lasers let you start small.

Others need to be big.

For one encampment - can use laser today.
Q:

Key leverage point for accelerating energy singularity.

Great question.

What is big barrier?

Low cost of non-renewable energy.

By using dirty energy without cleaning up. So abundant. Not a good economic incentive to do the investments in the technology. Doing it even though it doesn't pay off yet.
Q: Chiara: ...

Why not Thorium today?

Uranium was low hanging fruit.

Canada had a cheaper kind of Uranium available.

Burning Thorium is new. Inertia. Big companies at lead in US, built different kind of reactor.

Navies built pressurized water reactors.

When you look at it, evey country in the world has Thorium.

1:04 [applause]

Peter had inspiration to do XPrize from book about Lindbergh.

Gregg gave him the book.
At lunch with Gregg:
Ceramic Houses

http://en.wikipedia.org/wiki/Ceramic_houses

http://www.kheper.net/parameters/pages/ceramic_houses.html
Daniel Yergin, The Prize

http://en.wikipedia.org/wiki/Daniel_Yergin

http://www.amazon.com/Prize-Epic-Quest-Money-Power/dp/0671799320
Steve Jergeson?

Reverse Osmosis Energy Store

http://www.icrepq.com/icrepq'10/631-Salazar.pdf

http://en.wikipedia.org/wiki/Pressure-retarded_osmosis

Osmotic Power - how does it work?

http://www.statkraft.com/energy-sources/osmotic-power/default.aspx

http://spectrum.ieee.org/energy/renewables
Solar plane soars for 24 hours

By Eliane Engeler

The Associated Press

Experimental aircraft collects enough energy in daylight to stay up all night

http://www.usatoday.com/printedition/news/20100709/flight09_st.art.htm
2:07 Gregg's presentation continues in 583C:

2:20

- What is WORK.

In old days, had source of energy, had to locate business there.

Water wheels.

- Mechanical Power Transmission

Moved 10s of meters.

- How do we measure rate of energy (power)?

James Watt, inventor of steam engine

One horsepower - in old English system, moving 1 pound object 550 feet.

or a 550 pound object moved 1 foot in 1 second.

- Horsepower relates to cars now.

- One Kilowatt.

A watt is one joule per second.

One joule is lifting an apple up off the table.

One bright incadescent bulb is 100 watts. 10 of those is a kilowatt.

- 1 kilowatt=1.34 horsepower

http://en.wikipedia.org/wiki/Watt

http://en.wikipedia.org/wiki/Kilowatt_hour

-

-

-

- Nimitz aircraft carrier

http://en.wikipedia.org/wiki/Nimitz_class_aircraft_carrier

Has a billion and a half horse power nuclear reactor

- Kilowatt is a rate. For an hour, kW-hour - run 10, 100 Watt lightbulb for an hour

Amazing how cheap you can buy that. One coin from your pocket.
Gary: how many kilowatts in a joule?

http://en.wikipedia.org/wiki/Joule

Bryce: how many joules per horsepower?

1 joule / second
- Wouldn't it be great if ...

Move mechanical energy to somewhere else.

Enter Benjamin Franklin. My favorite. His sense of humor so bad they would not let him write the Dec of Inde because he would put puns in it.
- Not in my back yard.

\"NIMBY\"

I don't want that powerplant in my back yard.

-BANANA

Build Absolutely Nothing Anywhere Near Anyone
- Three main ways to get electrons moving

1/ Move a magnet near a wire

2/ battery. dissimilar

3/ solar cell - shine light on semiconductor junction Photo voltaics
- other ways

Heat junction - thermocouple. Radio-isotope Thermal Generator - used on all deep space probes.
Squeeze a crystal - piezoelectric - example: gas lighter makes spark.
Q: What kind of experiment left on moon?

That was power source to allow on-going experiments.

Passive laser retro reflector is all that is still working on moon.
-Princeton Piezo heel

Old radios used energy of 80 iPods. Tubes required 20 Watts each.
Marko: Charge wirelessly?
That kind of wireless energy - inductive.

This is not wireless - you'd need wire running up leg.

- Move a coil of wire in a magnetic field.

Take sensitive volt meter, just wave magnet around.

Note that wires here are moving and have contacts. That's a weak point.

Alternatively, Spin magnets instead of wire. That's what you do in your car.
Where do you get energy to turn the crank.

Number 1 way we get energy is heat difference, or waterfall.

- [aerial photo of coal fired power plant on Missouri river]

Power generator in my backyard.

Low sulfur coal.

You can see the big coal pile.

Trains - uni-trains. 120-130 cars, just piles of coal.

2.4 Gigawatts.

Back to the Future.

\"jigga\" or Gig-a-watt

What's the difference between a dead squirrel and a dead accordian player on the road?

It is possible the squirrel was on the way to a paid gig.
- Turning the crank with steam from burning coal.

Source of high temp.

Need also a source of low temperature.

- Turbine in Labadie Power Plant

Looks very much like a jet engine

[roar outside of passing jet]

Right on cue!
These are about 30% efficient converting heat to electrical energy.

70% goes up the stack.

Theoretically, higher.

Secret part of plant is the river. Providing cooling water.

Think of it as a waterfall.

How far does it drop.

In theory, to make it work best, to get perfect Carnot efficiency, the cold side would be at absolute zero.
http://en.wikipedia.org/wiki/Carnot_efficiency#Efficiency
- Baldwin Lake. 1.7 Gigawatt plant near Sparta, IL.

- US Sources of Elec.

Half coal

20% nuclear

20% nat gas

7% hydro

1.6% oil

Which country has most on a percentage basis? France 70%.
Chicago - gets to 30 below F.

You need power or you die when it is this cold.

Favorite solution - space solar power.

- The Power Grid - Electrical Magic?
- [photo of transmission pylons]

You have losses. Mostly as heat.

Whole grid is designed to minimize those loses.

Change characteristics of power. Change back and distribute to end users.

Like the way gears work.
- Current vs Voltage.

Use a water analogy.

You are familiar with static electricity.

Van de Graaf - high voltage, low current

http://en.wikipedia.org/wiki/Van_de_Graaff_generator
Current - amp

Pressure of electricity is measured in volts
River, moving slowly, moving lots of water.
High voltage, high current powerline will turn you into charred carbon and smoke.
- Names of Electrical Terms
http://en.wikipedia.org/wiki/Volt

http://en.wikipedia.org/wiki/Ampere

http://en.wikipedia.org/wiki/Watt
- Transformers: Electric gears
Think of magnetic field like blowing up a balloon.

When you make one grow on one coil, it can be taken out
- The Real Transformers
Resistance loss depending on current flow.

Jack up voltage, minimze current.
-

- Grid-a major \"public good\"

This month's National Geographic article. Mostly about future grids.

186,500 miles (300,000 km of grid)
The 21st Century Grid

Can we fix the infrastructure that powers our lives?

By Joel Achenbach

Photograph by Joe McNally

http://ngm.nationalgeographic.com/2010/07/power-grid/achenbach-text
There are times when your electricity may come from 1000 miles away.
- Overview of the Grid

You have to overbuild.

Getting more and more expensive.
- Storage. Not impossible

-

- Taum Sauk Pumped Storage

http://en.wikipedia.org/wiki/Taum_Sauk_Hydroelectric_Power_Station

- [photos]

Amerind UE

Important asset to them.

Paid a lot of money to fix damage.
- Consequences of lack of storage.

Utilities are not evil. They are owned by rate payers.

Intermittent sources are hard to integrate.
- Full circle - steam kettle to kettle.

How efficient from 1800's device? Very!

This machine wants heat. So you get good results.

- table of things and their power use

Microwave Oven - 1500 watts, but not used for long.

Well pump - 1500 watts.

You can see lights dim in your house when you kick on your hair dryer.
- In houses.

Lighting is 10%

Water heating is pretty large 10%

Cooling of air 17%

Kitchen appliances 29%

People think, \"if i change lightbulbs\"
- In offices

Lighting is 44%.

Buildings - people.

Energy considered free. Big windows. not how you design buildings if energy costs a lot of money.

European hotel - corridor is dark. Infrared sensor turns light on.

No power in room until I put my key in the door.
- New use of electricity.

Moving information. Servers. Big users of electricity.

Spinning drives around? No. Cooling systems.

Some companies who have lots of servers starting to put server farms along Columbia River Valley.

At speed of light, you don't care.

Lots of pipes for information, not so much for electricity.

All servers in the world use about same energy as the Netherlands.
- Light bulb. 19th century device.

You don't care, at the end of the day, until you get the bill.

What gets the job done?

10 % light, 90% heat.

LED is much better.

Cost - capital cost to acquire.

LED is simple in principle, fancy expensive factories to make.
Flourescent vs LED?

http://en.wikipedia.org/wiki/Fluorescent_lamp

http://en.wikipedia.org/wiki/LED_light
- Electricity

Some people will never switch, don't have money to make change.
Q: In Europe, cannot buy the old style bulb any more.

US has just passed a law, but no body knows it.

http://en.wikipedia.org/wiki/Phase-out_of_incandescent_light_bulbs

Energy Independence and Security Act of 2007 ?

The efficiency standards will start with 100-watt bulbs in January 2012 and end with 40-watt bulbs in January 2014.
Big reason people move to cities is to have power.

Rural electrification:

http://en.wikipedia.org/wiki/Rural_electrification

for light and laundry
Until people saw things made from lunar materials, they would not believe it.
Until people saw \"radio flashlight\" they didn't believe it.
http://en.wikipedia.org/wiki/Nikola_Tesla

3:31

http://www.johnmcphee.com/curve.htm

http://www.amazon.com/Curve-Binding-Energy-Alarming-Theodore/dp/0374515980

\"Ted Taylor is a theoretical physicist who was for many years a conceptual designer of atomic bombs. At Los Alamos Scientific Laboratory, he conceived and designed the largest-yield fission bomb ever exploded by any nation. Another of his bombs was, in its time, the lightest and smallest ever made.

Taylor later became the leader of a secret scientific effort, financed by the federal government, to make a spaceship the size of a sixteen-story building. The ship was of his invention and was to be called Orion. Powered by two thousand atomic bombs, exploding one at a time, it would move very rapidly to Mars, Jupiter, Saturn, and Pluto. The Limited Nuclear Test Ban Treaty of 1963 ended the project (but if human beings ever achieve travel much beyond the moon, some such vehicle will carry them).\"
Three Gorges Dam

http://en.wikipedia.org/wiki/Three_Gorges_Dam

Just over 4 solar power satellites.
3:46

Go over books one more time:

2081

http://en.wikipedia.org/wiki/2081:_A_Hopeful_View_of_the_Human_Future
New Yorker Magazine, Elizabeth Kobert

http://www.newyorker.com/reporting/2008/07/07/080707fa_fact_kolbert
A REPORTER AT LARGETHE ISLAND IN THE WINDA Danish community\u2019s victory over carbon emissions.by Elizabeth KolbertJULY 7, 2008

TEXT SIZE:SMALL TEXTMEDIUM TEXTLARGE TEXT

PRINT E-MAIL FEEDS
Once people on Sams\u00f8 started thinking about energy, a local farmer explains, \u201cit became a kind of sport.\u201d Photograph by Joachim Ladefoged.

RELATED LINKSSlide Show: Photographs of the Danish island of Sams\u00f8, by Joachim Ladefoged.

KEYWORDSClimate Change; Global Warming; Sams\u00f8;Denmark; Tranberg, J\u00f8rgen; Energy Reform;2,000-Watt Society

J\u00f8rgen Tranberg is a farmer who lives on the Danish island of Sams\u00f8. He is a beefy man with a mop of brown hair and an unpredictable sense of humor. When I arrived at his house, one gray morning this spring, he was sitting in his kitchen, smoking a cigarette and watching grainy images on a black-and-white TV. The images turned out to be closed-circuit shots from his barn. One of his cows, he told me, was about to give birth, and he was keeping an eye on her. We talked for a few minutes, and then, laughing, he asked me if I wanted to climb his wind turbine. I was pretty sure I didn\u2019t, but I said yes anyway.

We got into Tranberg\u2019s car and bounced along a rutted dirt road. The turbine loomed up in front of us. When we reached it, Tranberg stubbed out his cigarette and opened a small door in the base of the tower. Inside were eight ladders, each about twenty feet tall, attached one above the other. We started up, and were soon huffing. Above the last ladder, there was a trapdoor, which led to a sort of engine room. We scrambled into it, at which point we were standing on top of the generator. Tranberg pressed a button, and the roof slid open to reveal the gray sky and a patchwork of green and brown fields stretching toward the sea. He pressed another button. The rotors, which he had switched off during our climb, started to turn, at first sluggishly and then much more rapidly. It felt as if we were about to take off. I\u2019d like to say the feeling was exhilarating; in fact, I found it sickening. Tranberg looked at me and started to laugh.

Sams\u00f8, which is roughly the size of Nantucket, sits in what\u2019s known as the Kattegat, an arm of the North Sea. The island is bulgy in the south and narrows to a bladelike point in the north, so that on a map it looks a bit like a woman\u2019s torso and a bit like a meat cleaver. It has twenty-two villages that hug the narrow streets; out back are fields where farmers grow potatoes and wheat and strawberries. Thanks to Denmark\u2019s peculiar geography, Sams\u00f8 is smack in the center of the country and, at the same time, in the middle of nowhere.

For the past decade or so, Sams\u00f8 has been the site of an unlikely social movement. When it began, in the late nineteen-nineties, the island\u2019s forty-three hundred inhabitants had what might be described as a conventional attitude toward energy: as long as it continued to arrive, they weren\u2019t much interested in it. Most Samsingers heated their houses with oil, which was brought in on tankers. They used electricity imported from the mainland via cable, much of which was generated by burning coal. As a result, each Samsinger put into the atmosphere, on average, nearly eleven tons of carbon dioxide annually.
*FROM THE ISSUE

*CARTOON BANK

*E-MAIL THIS

Then, quite deliberately, the residents of the island set about changing this. They formed energy co\u00f6peratives and organized seminars on wind power. They removed their furnaces and replaced them with heat pumps. By 2001, fossil-fuel use on Sams\u00f8 had been cut in half. By 2003, instead of importing electricity, the island was exporting it, and by 2005 it was producing from renewable sources more energy than it was using.

The residents of Sams\u00f8 that I spoke to were clearly proud of their accomplishment. All the same, they insisted on their ordinariness. They were, they noted, not wealthy, nor were they especially well educated or idealistic. They weren\u2019t even terribly adventuresome. \u201cWe are a conservative farming community\u201d is how one Samsinger put it. \u201cWe are only normal people,\u201d Tranberg told me. \u201cWe are not some special people.\u201d

his year, the world is expected to burn through some thirty-one billion barrels of oil, six billion tons of coal, and a hundred trillion cubic feet of natural gas. The combustion of these fossil fuels will produce, in aggregate, some four hundred quadrillion B.T.U.s of energy. It will also yield around thirty billion tons of carbon dioxide. Next year, global consumption of fossil fuels is expected to grow by about two per cent, meaning that emissions will rise by more than half a billion tons, and the following year consumption is expected to grow by yet another two per cent.

When carbon dioxide is released into the air, about a third ends up, in relatively short order, in the oceans. (CO2 dissolves in water to form a weak acid; this is the cause of the phenomenon known as \u201cocean acidification.\u201d) A quarter is absorbed by terrestrial ecosystems\u2014no one is quite sure exactly how or where\u2014and the rest remains in the atmosphere. If current trends in emissions continue, then sometime within the next four or five decades the chemistry of the oceans will have been altered to such a degree that many marine organisms\u2014including reef-building corals\u2014will be pushed toward extinction. Meanwhile, atmospheric CO2 levels are projected to reach five hundred and fifty parts per million\u2014twice pre-industrial levels\u2014virtually guaranteeing an eventual global temperature increase of three or more degrees. The consequences of this warming are difficult to predict in detail, but even broad, conservative estimates are terrifying: at least fifteen and possibly as many as thirty per cent of the planet\u2019s plant and animal species will be threatened; sea levels will rise by several feet; yields of crops like wheat and corn will decline significantly in a number of areas where they are now grown as staples; regions that depend on glacial runoff or seasonal snowmelt\u2014currently home to more than a billion people\u2014will face severe water shortages; and what now counts as a hundred-year drought will occur in some parts of the world as frequently as once a decade.

Today, with CO2 levels at three hundred and eighty-five parts per million, the disruptive impacts of climate change are already apparent. The Arctic ice cap, which has shrunk by half since the nineteen-fifties, is melting at an annual rate of twenty-four thousand square miles, meaning that an expanse of ice the size of West Virginia is disappearing each year. Over the past ten years, forests covering a hundred and fifty million acres in the United States and Canada have died from warming-related beetle infestations. It is believed that rising temperatures are contributing to the growing number of international refugees\u2014\u201cClimate change is today one of the main drivers of forced displacement,\u201d the United Nations\u2019 high commissioner for refugees, Ant\u00f3nio Guterres, said recently\u2014and to armed conflict: some experts see a link between the fighting in Darfur, which has claimed as many as three hundred thousand lives, and changes in rainfall patterns in equatorial Africa.

\u201cIf we keep going down this path, the Darfur crisis will be only one crisis among dozens of others,\u201d President Nicolas Sarkozy, of France, told a meeting of world leaders in April. The Secretary-General of the United Nations, Ban Ki-moon, has called climate change \u201cthe defining challenge of our age.\u201d

In the context of this challenge, Sams\u00f8\u2019s accomplishments could be seen as trivial. Certainly, in numerical terms they don\u2019t amount to much: all the island\u2019s avoided emissions of the past ten years are overwhelmed by the CO2 that a single coal-fired power plant will emit in the next three weeks, and China is building new coal-fired plants at the rate of roughly four a month. But it is also in this context that the island\u2019s efforts are most significant. Sams\u00f8 transformed its energy systems in a single decade. Its experience suggests how the carbon problem, as huge as it is, could be dealt with, if we were willing to try.

ams\u00f8 set out to reinvent itself thanks to a series of decisions that it had relatively little to do with. The first was made by the Danish Ministry of Environment and Energy in 1997. The ministry, looking for ways to promote innovation, decided to sponsor a renewable-energy contest. In order to enter, a community had to submit a plan showing how it could wean itself off fossil fuels. An engineer who didn\u2019t actually live on Sams\u00f8 thought the island would make a good candidate. In consultation with Sams\u00f8\u2019s mayor, he drew up a plan and submitted it. When it was announced that Sams\u00f8 had won, the general reaction among residents was puzzlement. \u201cI had to listen twice before I believed it,\u201d one farmer told me.

The brief surge of interest that followed the announcement soon dissipated. Besides its designation as Denmark\u2019s \u201crenewable-energy island,\u201d Sams\u00f8 received basically nothing\u2014no prize money or special tax breaks, or even government assistance. One of the few people on the island to think the project was worth pursuing was S\u00f8ren Hermansen.

Hermansen, who is now forty-nine, is a trim man with close-cropped hair, ruddy cheeks, and dark-blue eyes. He was born on Sams\u00f8 and, save for a few stints away, to travel and go to university, has lived there his entire life. His father was a farmer who grew, among other things, beets and parsley. Hermansen, too, tried his hand at farming\u2014he took over the family\u2019s hundred acres when his father retired\u2014but he discovered he wasn\u2019t suited to it. \u201cI like to talk, and vegetables don\u2019t respond,\u201d he told me. He leased his fields to a neighbor and got a job teaching environmental studies at a local boarding school. Hermansen found the renewable-energy-island concept intriguing. When some federal money was found to fund a single staff position, he became the project\u2019s first employee.

For months, which stretched into years, not much happened. \u201cThere was this conservative hesitating, waiting for the neighbor to do the move,\u201d Hermansen recalled. \u201cI know the community and I know this is what usually happens.\u201d Rather than working against the islanders\u2019 tendency to look to one another, Hermansen tried to work with it.

\u201cOne reason to live here can be social relations,\u201d he said. \u201cThis renewable-energy project could be a new kind of social relation, and we used that.\u201d Whenever there was a meeting to discuss a local issue\u2014any local issue\u2014Hermansen attended and made his pitch. He asked Samsingers to think about what it would be like to work together on something they could all be proud of. Occasionally, he brought free beer along to the discussions. Meanwhile, he began trying to enlist the support of the island\u2019s opinion leaders. \u201cThis is where the hard work starts, convincing the first movers to be active,\u201d he said. Eventually, much as Hermansen had hoped, the social dynamic that had stalled the project began to work in its favor. As more people got involved, that prompted others to do so. After a while, enough Samsingers were participating that participation became the norm.

\u201cPeople on Sams\u00f8 started thinking about energy,\u201d Ingvar J\u00f8rgensen, a farmer who heats his house with solar hot water and a straw-burning furnace, told me. \u201cIt became a kind of sport.\u201d

\u201cIt\u2019s exciting to be a part of this,\u201d Brian Kj\u00e6r, an electrician who installed a small-scale turbine in his back yard, said. Kj\u00e6r\u2019s turbine, which is seventy-two feet tall, generates more current than his family of three can use, and also more than the power lines leading away from his house can handle, so he uses the excess to heat water, which he stores in a tank that he rigged up in his garage. He told me that one day he would like to use the leftover electricity to produce hydrogen, which could potentially run a fuel-cell car.

\u201cS\u00f8ren, he has talked again and again, and slowly it\u2019s spread to a lot of people,\u201d he said.

ince becoming the \u201crenewable energy island,\u201d Sams\u00f8 has increasingly found itself an object of study. Researchers often travel great distances to get there, a fact that is not without its own irony. The day after I arrived, from New York via Copenhagen, a group of professors from the University of Toyama, in Japan, came to look around. They had arranged a tour with Hermansen, and he invited me to tag along. We headed off to meet the group in his electric Citro\u00ebn, which is painted blue with white puffy clouds on the doors. It was a drizzly day, and when we got to the dock the water was choppy. Hermansen commiserated with the Japanese, who had just disembarked from the swaying ferry; then we all boarded a bus.

Our first stop was a hillside with a panoramic view of the island. Several wind turbines exactly like the one I had climbed with Tranberg were whooshing nearby. In the wet and the gray, they were the only things stirring. Off in the distance, the silent fields gave way to the Kattegat, where another group of turbines could be seen, arranged in a soldierly line in the water.

All told, Sams\u00f8 has eleven large land-based turbines. (It has about a dozen additional micro-turbines.) This is a lot of turbines for a relatively small number of people, and the ratio is critical to Sams\u00f8\u2019s success, as is the fact that the wind off the Kattegat blows pretty much continuously; flags on Sams\u00f8, I noticed, do not wave\u2014they stick straight out, as in children\u2019s drawings. Hermansen told us that the land-based turbines are a hundred and fifty feet tall, with rotors that are eighty feet long. Together, they produce some twenty-six million kilowatt-hours a year, which is just about enough to meet all the island\u2019s demands for electricity. (This is true in an arithmetic sense; as a practical matter, Sams\u00f8\u2019s production of electricity and its needs fluctuate, so that sometimes it is feeding power into the grid and sometimes it is drawing power from it.) The offshore turbines, meanwhile, are even taller\u2014a hundred and ninety-five feet high, with rotors that extend a hundred and twenty feet. A single offshore turbine generates roughly eight million kilowatt-hours of electricity a year, which, at Danish rates of energy use, is enough to satisfy the needs of some two thousand homes. The offshore turbines\u2014there are ten of them\u2014were erected to compensate for Sams\u00f8\u2019s continuing use of fossil fuels in its cars, trucks, and ferries. Their combined output, of around eighty million kilowatt-hours a year, provides the energy equivalent of all the gasoline and diesel oil consumed on the island, and then some; in aggregate, Sams\u00f8 generates about ten per cent more power than it consumes.

\u201cWhen we started, in 1997, nobody expected this to happen,\u201d Hermansen told the group. \u201cWhen we talked to local people, they said, Yes, come on, maybe in your dreams.\u201d Each land-based turbine cost the equivalent of eight hundred and fifty thousand dollars. Each offshore turbine cost around three million dollars. Some of Sams\u00f8\u2019s turbines were erected by a single investor, like Tranberg; others were purchased collectively. At least four hundred and fifty island residents own shares in the onshore turbines, and a roughly equal number own shares in those offshore. Shareholders, who also include many non-residents, receive annual dividend checks based on the prevailing price of electricity and how much their turbine has generated.

\u201cIf I\u2019m reduced to being a customer, then if I like something I buy it, and if I don\u2019t like it I don\u2019t buy it,\u201d Hermansen said. \u201cBut I don\u2019t care about the production. We care about the production, because we own the wind turbines. Every time they turn around, it means money in the bank. And, being part of it, we also feel responsible.\u201d Thanks to a policy put in place by Denmark\u2019s government in the late nineteen-nineties, utilities are required to offer ten-year fixed-rate contracts for wind power that they can sell to customers elsewhere. Under the terms of these contracts, a turbine should\u2014barring mishap\u2014repay a shareholder\u2019s initial investment in about eight years.

From the hillside, we headed to the town of Ballen. There we stopped at a red shed-shaped building made out of corrugated metal. Inside, enormous bales of straw were stacked against the walls. Hermansen explained that the building was a district heating plant that had been designed to run on biomass. The bales, each representing the equivalent of fifty gallons of oil, would be fed into a furnace, where water would be heated to a hundred and fifty-eight degrees. This hot water would then be piped underground to two hundred and sixty houses in Ballen and in the neighboring town of Brundby. In this way, the energy of the straw burned at the plant would be transferred to the homes, where it could be used to provide heat and hot water.

Sams\u00f8 has two other district heating plants that burn straw\u2014one in Tranebjerg, the other in Onsbjerg\u2014and also a district plant, in Nordby, that burns wood chips. When we visited the Nordby plant, later that afternoon, it was filled with what looked like mulch. (The place smelled like a potting shed.) Out back was a field covered in rows of solar panels, which provide additional hot water when the sun is shining. Between the rows, sheep with long black faces were munching on the grass. The Japanese researchers pulled out their cameras as the sheep snuffled toward them, expectantly.

Of course, burning straw or wood, like burning fossil fuels, produces CO2. The key distinction is that while fossil fuels release carbon that otherwise would have remained sequestered, biomass releases carbon that would have entered the atmosphere anyway, through decomposition. As long as biomass regrows, the CO2 released in its combustion should be reabsorbed, meaning that the cycle is\u2014or at least can be\u2014carbon neutral. The wood chips used in the Nordby plant come from fallen trees that previously would have been left to rot. The straw for the Ballen-Brundby plant comes mainly from wheat stalks that would previously have been burned in the fields. Together, the biomass heating plants prevent the release of some twenty-seven hundred tons of carbon dioxide a year.

In addition to biomass, Sams\u00f8 is experimenting on a modest scale with biofuels: a handful of farmers have converted their cars and tractors to run on canola oil. We stopped to visit one such farmer, who grows his own seeds, presses his own oil, and feeds the leftover mash to his cows. The farmer couldn\u2019t be located, so Hermansen started up the press himself. He stuck a finger under the spout, then popped it into his mouth. \u201cThe oil is very good,\u201d he announced. \u201cYou can use it in your car, and you can use it on your salad.\u201d

After the tour, I went back with Hermansen to his office, in a building known as the Energiakademi. The academy, which looks like a Bauhaus interpretation of a barn, is covered with photovoltaic cells and insulated with shredded newspapers. It is supposed to serve as a sort of interpretive center, though when I visited, the place was so new that the rooms were mostly empty. Some high-school students were kneeling on the floor, trying to put together a miniature turbine.

I asked Hermansen whether there were any projects that hadn\u2019t worked out. He listed several, including a plan to use natural gas produced from cow manure and an experiment with electric cars that failed when one of the demonstration vehicles spent most of the year in the shop. The biggest disappointment, though, had to do with consumption.

\u201cWe made several programs for energy savings,\u201d he told me. \u201cBut people are acting\u2014what do you call it?\u2014irresponsibly. They behave like monkeys.\u201d For example, families that insulated their homes better also tended to heat more rooms, \u201cso we ended up with zero.\u201d Essentially, he said, energy use on the island has remained constant for the past decade.

I asked why he thought the renewable-energy-island effort had got as far as it did. He said he wasn\u2019t sure, because different people had had different motives for participating. \u201cFrom the very egoistic to the more over-all perspective, I think we had all kinds of reasons.\u201d

Finally, I asked what he thought other communities might take from Sams\u00f8\u2019s experience.

\u201cWe always hear that we should think globally and act locally,\u201d he said. \u201cI understand what that means\u2014I think we as a nation should be part of the global consciousness. But each individual cannot be part of that. So \u2018Think locally, act locally\u2019 is the key message for us.\u201d

\u201cThere\u2019s this wish for showcases,\u201d he added. \u201cWhen we are selected to be the showcase for Denmark, I feel ashamed that Denmark doesn\u2019t produce anything bigger than that. But I feel proud because we are the showcase. So I did my job, and my colleagues did their job, and so did the people of Sams\u00f8.\u201d

round the same time that Sams\u00f8 was designated Denmark\u2019s renewable-energy island, a group of Swiss scientists who were working on similar issues performed a thought experiment. The scientists, all of whom were affiliated with the Swiss Federal Institute of Technology, asked themselves what level of energy use would be sustainable, not just for an island or a small European nation but for the entire world. The answer they came up with\u2014two thousand watts per person\u2014furnished the name for a new project: the 2,000-Watt Society.

\u201cWhat it\u2019s important, I think, to know is that the 2,000-Watt Society is not a program of hard life,\u201d the director of the project, Roland Stulz, told me when I went to speak to him at his office, in the Zurich suburb of D\u00fcbendorf. \u201cIt is not what we call G\u00fcrtel enger schnallen\u201d\u2014belt tightening\u2014\u201cit\u2019s not starving, it\u2019s not having less comfort or fun. It\u2019s a creative approach to the future.\u201d

Stulz, who is sixty-three, is a softspoken man with dark wavy hair and a salt-and-pepper mustache. He was trained as an architect and later became interested in energy-efficient building. In 2001, when he took over the 2,000-Watt Society, his mandate was to push it into the realm of the practical. (His work is funded in part by the Swiss Federal Institute of Technology, which has campuses in Zurich and Lausanne, and in part by private donations.) He began holding meetings that brought researchers together with government officials from cities like Zurich and Basel.

\u201cI divided them into groups,\u201d Stulz recalled. \u201cAnd I told them, At four o\u2019clock each group must come and tell the whole session what project they will do in the future, and who will lead the projects. And they said, Oh, it\u2019s not possible. But at four o\u2019clock everybody came with a project. And that\u2019s how we started.\u201d The cantons of Geneva and Basel-Stadt and the city of Zurich subsequently endorsed the aims of the 2,000-Watt Society, as did the Swiss Federal Department of the Environment, Transport, Energy, and Communications. \u201cAt first glance, the objective of a two-thousand-watt society appears unrealistic,\u201d Moritz Leuenberger, the head of the federal department, has said. \u201cBut the necessary technology already exists.\u201d

One afternoon, Stulz took me to visit the headquarters of an aquatic-research center known as EAWAG, which was designed to meet the 2,000-Watt Society\u2019s energy-efficiency goals. (EAWAG is an acronym for a German name so complicated that even German speakers can\u2019t remember it.) We drove over in his Volvo, which runs on compressed natural gas produced in part from rotting vegetables. When I first caught sight of the place, I thought it was covered with banners; these turned out to be tinted-glass panels. Inside, hanging from a set of chains in a large atrium, was what I took to be a sculpture of a bug. This turned out to be a model of a water molecule, enlarged some ten billion times.

Among the many unusual features of the EAWAG Center is a lack of usual features. The building, which opened in 2006, has no furnace; it is so tightly insulated that, on most days, the warmth thrown off by the office equipment and the two hundred people who work inside is enough to keep it comfortable. Additional heat is provided by the sun\u2014in winter, the outside panels tilt to allow in the maximum amount of light\u2014and by air sucked in from underground. The building also has no conventional air-conditioners: in summer, the panels tilt to provide shade, and if the building gets hot during the day, at night the windows at the top of the atrium open, and the warm air rushes out. It supplies about a third of its own electricity with photovoltaic panels installed on the roof, and gets its hot water from solar collectors. Its bathrooms are equipped with specially designed \u201cno mix\u201d toilets that separate out urine, which contains potentially useful phosphorus and nitrogen. (\u201cExploiting common waste as a resource is a mark of sustainable civilization,\u201d a booklet on the building observes.)

\u201cIt\u2019s not a miracle, such a building,\u201d Stulz told me when we went to have a cup of coffee in the center\u2019s cheerfully modernist cafeteria. \u201cIt\u2019s just putting smart elements together in a smart way.\u201d Outside, it was rainy and forty-three degrees; inside the temperature was a pleasant seventy.

ne way to think about the 2,000-Watt Society is in terms of light bulbs. Let\u2019s say you turn on twenty lamps, each with a hundred-watt bulb. Together, the lamps will draw two thousand watts of power. Left on for a day, they will consume forty-eight kilowatt-hours of energy; left on for a year, they will consume seventeen thousand five hundred and twenty kilowatt-hours. A person living a two-thousand-watt life would consume in all his activities\u2014working, eating, travelling\u2014the same amount of energy as those twenty bulbs, or seventeen thousand five hundred and twenty kilowatt-hours annually.

Most of the people in the world today consume far less than this. The average Bangladeshi, for example, uses only about twenty-six hundred kilowatt-hours a year\u2014this figure includes all forms of energy, from electricity to transportation fuel\u2014which is the equivalent of using roughly three hundred watts continuously. The average Indian uses about eighty-seven hundred kilowatt-hours a year, making India a one-thousand-watt society, while the average Chinese uses about thirteen thousand kilowatt-hours a year, making China a fifteen-hundred-watt society.

Those of us who live in the industrialized world, by contrast, consume far more than two thousand watts. Switzerland, for instance, is a five-thousand-watt society. Most other Western European countries are six-thousand-watt societies; the United States and Canada run at twelve thousand watts. One of the founding principles of the 2,000-Watt Society is that this disparity is in itself unsustainable. \u201cIt\u2019s a basic matter of fairness\u201d is how Stulz put it to me. But increasing energy use in developing countries to match that of industrialized nations would be unacceptable on ecological grounds. Were per-capita demand in the developing world to reach current European levels, global energy consumption would more than double, and were it to rise to the American level, global energy consumption would more than triple. The 2,000-Watt Society gives industrialized countries a target for cutting energy use at the same time that it sets a limit for growth in developing nations.

The last time Switzerland was a two-thousand-watt society was in the early nineteen-sixties. By the end of that decade, energy use had reached three thousand watts, and by the mid-seventies it was up to four thousand watts. This rapid rise could be said to follow from technological advances\u2014the spread of automobiles, the advent of jet travel, the proliferation of appliances and electronic devices\u2014or it could be seen as just the reverse: a failure to apply technology where it is needed. A few years ago, a group of Swiss scientists published a white paper\u2014or, to use the Swiss term, a \u201cwhite book\u201d\u2014on the feasibility of a 2,000-Watt Society. Relying on widely agreed-upon figures, the scientists estimated that two-thirds of all the primary energy consumed in the world today is wasted, mostly in the form of heat that nobody wants or uses. (\u201cPrimary energy\u201d is the energy contained in, say, a lump of coal; \u201cuseful energy\u201d is the light emitted by a bulb once that coal has been burned to produce steam, the steam has been used to run a turbine, and the resulting electricity has been transmitted over the grid to heat the bulb\u2019s filament.) This same paper concluded that, with currently available technologies, buildings could be made eighty per cent more efficient, cars fifty per cent more efficient, and motors twenty-five per cent more efficient.

In Switzerland, I visited several other buildings that, like the EAWAG Center, had been specifically designed to maximize efficiency. One was an upscale apartment building in Basel. The apartments have eighteen-inch-thick walls filled with insulation, triple-paned windows coated with a special reflective film, and a heat-recovery system that captures eighty per cent of the energy normally lost through ventilation. Instead of a boiler, it has a geothermal heat pump, which essentially sucks energy out of the groundwater. In the summer, the same system is used for cooling. (In compliance with Swiss building codes, the building also contains a bomb shelter.)

\u201cThe construction industry is very traditional,\u201d Franco Fregnan, an engineer who showed me around the apartments, said. \u201cIf you bring an innovation to them, you usually have to wait another generation until it arrives into a building. And we are trying to change that, step by step.\u201d

\u201cIt usually makes sense to become more intelligent in any human activity,\u201d Stulz told me. \u201cAs the former Saudi Arabian oil minister Sheikh Yamani once said, the Stone Age didn\u2019t end because there were no more stones. It ended because people became more intelligent. \u201d

hat would it take to lead a two-thousand-watt life? When I posed this question to Stulz, he gave me another research paper, which offers case studies of six fictionalized households. The Jeannerets are an imaginary family of four who live in Glattbrugg, a town north of Zurich. They own an energy-efficient house, travel by electric bike or train, and occasionally rent a car\u2014they belong to a car-sharing service\u2014to do their grocery shopping. The Moeris, fictional farmers who live northeast of Bern, generate their own electricity with natural gas produced from cow manure; and Alain, Michel, Angela, and Marl\u00e8ne, fictional students living in Geneva, share all their appliances, use the tram, and like to go hiking in the French Alps during school breaks. \u201cThere is no formula for how to achieve a two-thousand-watt society,\u201d the paper declares. \u201cThree things are needed: societal decisions. . . technical innovation, and the resolve of every individual to act in an energy-conscious way.\u201d

Very broadly speaking, the average Swiss today uses energy as follows: fifteen hundred watts per day for living and office space (this includes heat and hot water), eleven hundred watts for food and consumer items (the energy that it takes to produce and transport goods is referred to as \u201cembodied\u201d or \u201cgray\u201d energy), six hundred watts for electricity, five hundred watts for automobile travel, two hundred and fifty watts for air travel, and a hundred and fifty watts for public transportation. Each person\u2019s share of Switzerland\u2019s public infrastructure, which includes facilities like water- and sewage-treatment plants, comes to nine hundred watts. Reducing these five thousand watts to two thousand would seem to require a significant reduction in every realm. Assuming that infrastructure-related consumption could be cut to five hundred watts, a person who continued to use fifteen hundred watts for living and office space would have nothing left for food, electricity, and transportation. Similarly, a person who continued to travel and use electricity at current rates would consume two thousand watts without having anywhere to live or work, or anything to eat.

While I was in Switzerland, I kept looking for people who actually led two-thousand-watt lives.

\u201cI\u2019m pretty close, except for this stupid air travel,\u201d Gerhard Schmitt, the vice-president for planning and logistics at the Zurich campus of the Swiss Federal Institute of Technology, told me. \u201cI go once to Shanghai and it\u2019s gone.\u201d (A round-trip flight between Zurich and Shanghai is the equivalent of using something like eight hundred watts continuously for a year.)

\u201cLet\u2019s skip that question,\u201d Stulz said when I put it to him. While he lives in an energy-efficient apartment, he, too, travels a great deal; when I visited, he had just returned from a conference in New Delhi, a round trip that used roughly the equivalent of six hundred watts for the year.

The one person I spoke to who did seem to be leading a two-thousand-watt life, or something very near to it, was an engineer named Robert Uetz. Uetz works in the same building as Stulz, and when we returned from visiting the EAWAG Center he was still in his office, even though it was after six. Stulz encouraged me to go talk to him.

\u201cWe don\u2019t experience it as a restriction,\u201d Uetz told me of his two-thousand-watt life style. \u201cOn the contrary. I don\u2019t feel that we\u2019re giving up anything.\u201d Uetz and his wife, a dentist, live with their two children in the city of Winterthur, near Zurich. About ten years ago, they bought a two-thousand-square-foot house in a newly built energy-efficient development. The house is heated with a geothermal heat pump\u2014\u201cIt\u2019s crazy to heat a house with fossil fuels,\u201d Uetz said\u2014and has a solar hot-water system. Uetz added photovoltaic panels to the roof to produce electricity; in the winter the panels produce somewhat less power than the house uses\u2014it\u2019s equipped with the most energy-efficient lights and appliances the family could find\u2014and in the summer they produce somewhat more, so that over the course of the year the house\u2019s electricity use nets out to zero.

\u201cThe most important decision was that we wouldn\u2019t have a car,\u201d Uetz told me. \u201cThat was a conscious decision. We looked for a house where we didn\u2019t need a car.\u201d Driving a lot\u2014even in what, by today\u2019s standards at least, counts as an energy-efficient vehicle\u2014also makes it difficult to live within two thousand watts. A person who drives a Toyota Prius ten thousand miles a year consumes roughly two hundred and twenty-five gallons of gasoline. This is equivalent to consuming around eight thousand kilowatt-hours, or to using nearly a thousand watts on a continuous basis. (For a family of four, the same gasoline consumption would come to almost two hundred and fifty watts per person.)

\u201cIt\u2019s a matter of what you\u2019re used to, but I find taking the train a lot more pleasant than driving,\u201d Uetz went on. \u201cOn the train I can work and relax. If I took a car, I\u2019d have to worry about parking and traffic, rain, snow, and a certain number of people who can\u2019t drive but are on the road anyway.\u201d When Uetz and his family go on vacation, they travel by rail. \u201cThe only thing I\u2019d say that is sort of a restriction is the flying,\u201d he said. \u201cBecause, obviously, with the train where you can go is limited. We can\u2019t go to China, or if we did it would take a week.\u201d

\u201cI don\u2019t make a religion out of it,\u201d he added. \u201cI wouldn\u2019t do it if I didn\u2019t feel good about it\u2014it\u2019s how I like to live.\u201d

y the 2,000-Watt Society\u2019s own reckoning, cutting consumption is just half\u2014or, perhaps more accurately, a quarter\u2014of what needs to be done. The project\u2019s ultimate goal is a world where people consume no more than two thousand watts apiece and where fifteen hundred of those watts come from carbon-free sources. In such a world, everyone would use energy sparingly, like Robert Uetz, and generate it renewably, like J\u00f8rgen Tranberg. In such a world, filled with windmills and net-zero houses, carbon emissions would fall sharply, and the concentration of CO2 in the atmosphere would slowly level off. But how realistic is such a scenario?

Before I left Switzerland to fly back to New York (a trip equivalent to using roughly two hundred and fifty watts continuously for a year), I went to speak to the president of the research council of the Swiss National Science Foundation, Dieter Imboden. Imboden, who is sixty-four, is a compact man with an oval face and silvery hair. He received his training in theoretical solid-state physics, later became interested in environmental physics, and for several years chaired the Swiss Federal Institute of Technology\u2019s environmental-sciences department. In the late nineties, he served as the director of the 2,000-Watt Society. He said that as a scientist he could see no technical barriers to creating a two-thousand-watt world.

\u201cWe are putting our mental energy into the wrong basket,\u201d he told me. \u201cNothing has to be reinvented\u2014for an engineer it\u2019s not even a challenge.\u201d

\u201cThe problems of the twenty-first century are a different kind of problem,\u201d he went on. \u201cAnd I think our society will be measured according to the solution of this new kind of problem, which cannot be solved with the same recipe as the flight to the moon, or the Manhattan Project. It\u2019s a qualitative difference\u2014a paradigm change in the role of science for our society.\u201d

He continued, \u201cThe difficult thing is what I call \u2018constructed Switzerland.\u2019 You in America could call it \u2018constructed United States\u2019\u2014the buildings and how they are built, but also where they are built and, even more important, the roads, the railroads, the lines for energy, for wastewater, and so on. It\u2019s not economically feasible to replace everything in one instant.\u201d But since infrastructure should in any case be replaced at the rate of roughly two per cent a year, if the project is approached incrementally, it\u2019s a different task. Then, Imboden said, \u201cit suddenly is feasible.\u201d

As of yet, no one has undertaken a rigorous analysis of the economics of a transition to two thousand watts. Researchers have tended, rather, to focus on the price of stabilizing carbon-dioxide levels in the atmosphere at a given concentration\u2014either, say, five hundred and fifty parts per million, which is double pre-industrial levels, or four hundred and fifty parts, which, many climate scientists say, is the very highest level advisable. Perhaps the most often cited economic study is the Stern Review, commissioned by the British government and named for its lead author, Sir Nicholas Stern, formerly the chief economist for the World Bank. The Stern Review, published in October, 2006, concluded that greenhouse-gas levels could be stabilized below double pre-industrial concentrations at a cost to global G.D.P. of around one per cent a year. (The Stern Review considered not just CO2 but other greenhouse gases, like methane and nitrous oxide, as well.) An analysis released last year by the Swedish utility Vattenfall, with research assistance from the American consulting firm McKinsey & Company, reached a similar conclusion: it determined that many measures to reduce carbon emissions, like improving building insulation, would save money, while others, like installing wind turbines, would carry a price. The Vattenfall report estimates that \u201cif all low-cost opportunities are addressed,\u201d CO2 levels could be stabilized at four hundred and fifty parts per million with an annual expenditure of six-tenths of one per cent of global G.D.P.

Though one per cent of the global economy is clearly a lot of money, in the grand scheme of things it\u2019s also clearly manageable. It is about a ninth of what\u2019s currently spent on health care, a seventh of what\u2019s spent on oil, and half of what\u2019s spent on defense. (More than forty per cent of all the world\u2019s military expenditures are made by the United States.) Perhaps most pertinent, it\u2019s a far smaller figure than the cost of inaction. The Stern Review projects that if current emissions trends are allowed to continue, the eventual damage from climate change will \u201cbe equivalent to losing at least 5% of global GDP each year, now and forever,\u201d and that \u201cif a wider range of risks and impacts is taken into account\u201d that figure could \u201crise to 20% of GDP or more.\u201d

Twenty years ago, NASA\u2019s chief climate scientist, James Hansen, testified on Capitol Hill about the dangers of global warming. Just a few days ago, Hansen returned to the Hill to testify again. \u201cNow, as then, frank assessment of scientific data yields conclusions that are shocking to the body politic,\u201d he said. \u201cNow, as then, I can assert that these conclusions have a certainty exceeding ninety-nine per cent. The difference is that now we have used up all slack in the schedule.\u201d Hansen went on to warn that there would be no practical way to prevent \u201cdisastrous\u201d climate change unless the next President and Congress act quickly to curb emissions. Few parts of the U.S. may be as windy as Sams\u00f8, or as well organized as Switzerland, but just about everywhere there are possibilities for generating energy more inventively and using it more intelligently. Realizing these possibilities will require a great deal of effort. We may well decide not to make this effort. Such a choice to put off change, however, will merely drive us toward it. \u2666

Read more: http://www.newyorker.com/reporting/2008/07/07/080707fa_fact_kolbert?currentPage=all#ixzz0tK3ZtsDl
=============== DURING-presentation notes


Can't stand the colors of Etherpad? Turn them off with Pad Options.

(Note: this changes the color for EVERYONE.)
Color card definitions:

Green - Agree

Red - Disagree

Yellow - Slow down (content) / I don't understand

Grey - Speed up we know this stuff / Move on to another topic
hands moving apart (sideways) - speak slower, speak more clearly
Link to backchan.nl for questions:

http://singularity.backchan.nl/meetings/view/157 First speaker

http://singularity.backchan.nl/meetings/view/158 Second speaker

http://singularity.backchan.nl/meetings/view/159 Third speaker
GSP10 Book List:

http://bit.ly/GSP10BookList
=============== POST-presentation notes
Instant Evaluation:

https://spreadsheets.google.com/a/singularityu.org/viewform?formkey=dGVzY3JzR3VFeTJjSElVcjM3REFpY0E6MQ
Reminder: PASSWORD CHANGE on MONDAY 12 JULY.
"

http://su-etherpad.com/core-Jul20-EES

"7/20/2010

Tuesday

9 am - 12 pm

EES Session: Edge Issue, Conservation and Efficiency

B583C
9:20 am - 9:40 am: Daniel Kammen (http://kammen.berkeley.edu/ )

Designing a Climate Friendly Future

Slides: http://bit.ly/aNYb7L

9:00 am - 9:20 am: Sunil Paul (http://www.linkedin.com/in/sunilpaul )

Slides: http://bit.ly/drkuJw

9:40 am - 10:00 am: Rohit Aggarwala (http://www.worldleaders.columbia.edu/participants/rohit-t-aggarwala )

Slides: http://bit.ly/cWokjE

10:00 am - 10:15 am: Explanation of the wedges/simulation rules and split into 4 groups (corresponding groups A, B, C, D in rooms A, B, C, D)

10:15 am - 10:30 am: Coffee Break

10:30 am - 11:15 am: Groups work in separate rooms

11:15 am - 11:45 am: Groups reconvene in Central Room: Each Group presents

11:45 am - 12:00 m: Final Discussion and Awards Ceremony
Think about the most important variable to scaling key wedges in \u201cwedge analysis\u201d \u2013 how they think about what\u2019s necessary to actually take these to scale. Following the explanations of the rules, students will then break into four groups in Building 583C: the already pre-assigned rooms A, B, C, and D. Sunil, Dan Rohit will rotate among the different rooms and serve as general facilitators. The best teams will win special awards in a closing ceremony.
Three groups (A, B and C) will work for an hour on how exponential technology can ensure that wedges (from GigatonThrowdown Report or otherwise) reach the scale necessary to lighten the global emissions load. These three groups will break out to each address one of the following key wedges:

A. Solar PV (Room A)

B. Building Efficiency (Room B)

C. Biofuels (Room D)

These three groups should spend a small part of their hour checking the core assumptions from GigatonThrowdown (and other sources as appropriate, including other lectures given this summer) of how the particular wedge they\u2019re working on will get to scale. They should spend the majority of their time developing a story of how exponential hard, soft, social, and economic approaches can ensure and/or accelerate the growth of their chosen wedge. (See http://www.gigatonthrowdown.org ) Each group should designate a reporter who will summarize the group\u2019s story back to the plenary group after an hour, and the three workshop leaders will select a winning wedge that will get a special prize.
The other group (D) will go to Room D and engage in an hour-long carbon trading simulation. The winning team in this simulation will also receive a special prize. (See the simple rules of the game below).

_______________________________________________________________________________________________

Info for Groups A, B, C: Wedge Analysis and GigatonThrowdon Summary (read: http://www.gigatonthrowdown.org/files/Gigaton_Overview.pdf)

_______________________________________________________________________________________________
Info for Group D: The Art of a Deal: A Kyoto Protocol Simulation (read: http://www.sciencecases.org/kyoto_protocol/kyoto_protocol_notes.asp )
In this simulation, you will represent one of several countries attempting to barter an international agreement to lower global carbon dioxide (CO2) emissions. For simplicity, there will only be three countries that need to lower their CO2emissions (Countries A, B, and C). To add a bit of realism, there will be one developing nation (Country D) that is exempt from any reduction agreement.

The goal of each country is to maximize its \u201cscore\u201d as computed below. At the end of the simulation, each member of the highest-scoring country will have their names thrown into a hat for a $50 campus bookstore gift certificate.
Rules

Countries A, B, and C will begin the simulation with 100 CO2 units and $100 per country. Country D has 50CO2 units for sale and $20 to start.

An agreement must be reached that lowers global CO2 emissions of Countries A, B, and C by 100 CO2 units. The 100 could include CO2 purchased from Country D. If an agreement is not reached, no names go in the drawing for the gift certificate.

Countries A, B, and C must approve any agreement unanimously. Country D has no vote.

Each country will designate two representatives that may speak to any other country or address the countries as a whole.

Country D has 50 CO2 units available for sale. Any country can purchase these units with their dollars. These 50 units can be sold in smaller blocks to separate countries. Any money used to purchase CO2 units is given directly to Country D.Any CO2 unit sales or reductions must be done in whole units.

Scoring: Determining the Winner
Countries A, B, and C

At the beginning of the simulation, Countries A, B, and C all have 200 points ($100 + 100 CO2 units). However, at the end of the simulation if the CO2 units of A, B, or C are below 100, their economies suffer. If the CO2 emissions are below 100, the economy must go down by one-half the amount below 100 units. For example, if Country A ends up with 80 CO2 units and $100, their economy will end up at $90. Their total score will then be 170.

Country D

The final score will be three times the economic size in dollars. Your starting economy size is $20, for a starting score of 60. No credit will be given for unsold CO2 units.
Notes from video

http://www.gigatonthrowdown.org/video/

7:20 2 million more direct jobs from clean energy

7:44 To stabilize climate, we've got to reduce carbon by 5-7 gigatons per year

9:10 All fossil fuel use is a threat to our national security

9:50 Biggest impact is through fuels - cellulosic ethanol

11:51 Over the next 10 years, $13 T will be invested in fossil fuels worldwide

12:31 Three things needed

Capital - waiting

Technology - capable of scaling

Policy - remaining open question

13:10 Rules of the game are designed for the fossil fuel industry

13:50 Key is to have stable policy

14:10 Being subject to politics is \"death\"

14:50 [slide showing 93% / 73% / 77% annual drops in wind turbine demand as production tax credits expired in 2000 / 2002 / 2004] This has led to bankruptcy of every wind company in the US.
9:15 Jose: Final day of Phase I

Three speakers this morning.

Dan Kammen from Berkeley.
Designing a Climate Friendly Future

Daniel M. Kammen

Slides: http://bit.ly/aNYb7L
Dan: Thank you.

We have a heavy agenda.

Workshop part of this will be designing a clean, low carbon solution.

The world's largest problem.

We have an hour.

Among the policies, technolgies, investment areas

design a coherent plan.

To put limited resources and huge amount of needs in perspective.
Steve Schneider passed away.

Important about him - very rare. Academics don't normally take their research results and jump into the public battle.

Spoke the truth.

The number of public officials who speak the truth - is rare.
Two aspects:

The data - Amory Lovins used to say, \"In God we trust, all others bring data\"

Steve highlighted what the data say.

Just to highlight - dramatic retreat of ice in Himalayas.

Incredible loss of ice in Rockies, Andes.

In 2009, spring arrives a week earlier and fall ends a week later.

Dramatic change for all kinds of animal species.
People look at the model results.

Find simple, clear ways to say what is going on.

99.9% of climate scientists agree on the data.

Yet 50% of people ... leave the train. Don't believe the human causes.

\"Climate-gate\"

Real failure to communicate the data.

June is the warmest in human history.

Who did NOT see that result? [most]

What I will show you is the amount of sea ice at the summer minimum.

There will be noise in the record.

Snapshot.

1 million square km.

Two things going on in this record.

Model free assessment of where we are today.

Even if one data point is dramatically important.

Linear trend - 50% loss in sea ice.

Both this linear trend with noise, hint

the thing that climate scientists suspect

we know there are surprises lurking for us as the world warms.

Hard to tell the story of the absolute calamities coming
Goal. Devise a climate plan.

History of US GHG emissions.

US is 1/4. China 1/4.

GHG in billions of tons on this graph.

Missed Kyoto goal - no comment.
For all bad press of IPCC for Climategate.

Done relative to 1990 baseline - 20% reduction from today. Then 80% from there.

De-carbonize the entire economy in 40 years.

We know how to get here.

We do not know how to do this [drop to 2050]

Full path is not yet discovered.
Copenhagen - 25-40% below 1990.

CA goal, get back to 1990 level.

Bills in US, mid-term goals.
Client science is very clear.

Snow in N hemisphere.

Series of IPCC reports - first: might take a decade to get clear data, second balance of data, third most data, fourth significant risk - warming will impact global poor.

So that is supposed to depress you all.
We do have successes. Look at kWh/person.

Dramatic decoupling that contributed to CA.

Today CA is about 40% better than national avg.

NY, RI, Wisconsin have done above national avg.

Denmark is 40% better than CA.

Efficient use of waste materials.

kWh / person - while, per capita looks great, added a lot of capitas.
San Fran - all incandescent lighting.

Sunil will talk about the economics.

Hard to replace a brown electron with a green electron if there is not better quality service along with it.
Going green doesn't just get you brownie points in heaven, it saves you money or gives you better service.
Lurking problems in the greenhouse. Any Canadians? From Alberta? They have bitumen.

Used to be called tar sands. Now called oil sands.

These are the piles of waste sulfur.

Venezuela.

More oil in Alberta than in Saudi Arabia.

Takes about 30% of energy, natural gas bubbled through soil. venting H to atmosphere.

to get oil

So we pass through two cleaner fuels to get oil.
Gigaton Throwdown

Building another nuclear sector.

Thinking about how to grow these.

Solar - 1990 assessments. Large fields, distributed homes, off-grid. Made estimates.

200 absolute solar enthusiasts - this what we forecast. This is what happened [much higher growth]

We thought least cost $/W.

Kenya - as % of pop. - only 10-12 W. but small distributed systems at high cost grew much more quickly than central ones.

Now seeing German, Portugal, Ontario - large scale.

Experts can be dead wrong.

May be able to come up with interesting new ways - because we had our heads up ... something.
- Public R&D.

Manhattan

Apollo

Reagan Defence

Jose: Star Wars
War on Terror

6-7x increase in short term.

My group will talk about being in charge of that budget.

Think in broad terms about how to scale spending.

Rob Socolo/Princeton.

Play the wedge game.
Sectoral goal.

Need 7 wedges to get to ...
9:45 Sunil Paul [applause]

Instigator/ organizer of Gigaton Throwdown.

My primary role is as an industrialist

Spring Ventures

invest only in clean tech.

Reason I want to talk to you today -

got into this world of investing - the most powerful lever

to get change in climate and energy security.

As I continued to invest - a friend said,

\"You clean tech guys could make a bunch of money and not make a bit of difference.\"

Solar could increase by a factor of 10 or 20, but that would not matter.

Interest in this category - that realization.

What DOES it take to make a big difference in climate and security.

Jose said this would be a heavy session - a gigton. 1 billion tons.

What is that?

An enormously heavy, big.

Give some idea, more palpable realization of what is a billion tons.

Seems hard to believe that a gas,

something you breathe out every day,

could be so heavy.

A car emits 3-4x its own weight - a Prius for example - emits 3x its own weight in CO2.

If you had to make 1 gigatons - it would take 285 m Priuses.

Roughly entire US fleet. In that magnitude.

Problem is just enormous.

Reason we chose wedge is you need something you can get your head around.
What are the possible solutions?

: Efficiency, less waste.

: Solar
10:05

Major Bloomberg asked to start office for New York city.

Did not start out with the goal of being green.

Long term growth strategy for NY.
designing green results in energy efficiency as a byproduct
NYC has carbon footprint ~Portugal
\"best thing you can do to decrease your cabon footprint is to move to NYC\"
LA needs the least amount of energy to heat/cool
45% of NYC's space is in 2% of its buildings

VC challenge: Want a return of $800 million from $200 million. Can't change laws, can influence but can't dole out money and expect laws and regulations ot change. YOu have tremendous access to talent, brains, etc. Now where do you focus?
Organizing the fund, what are we focused on, what are we not focused on and why?
First:

You have a budget from the government.

What can you do?
Second:

You are a venture capitalist. Responsible for R&D.

Live by the constraints.

Can't change the law.

How do you get to maximum impact.
Third:

You are a public official.

Mayor Bloomberg.

You've got the political equation to deal with.
Challenge - how do you end up having the biggest impact.
10:46 Break

10:58
12:55 Jose:

Anniversary of moon landing

http://en.wikipedia.org/wiki/Apollo_11
4 cakes.

David Roberts' birthday.
"

http://su-etherpad.com/core-Jul7-BB-links

"LINKS FOR SESSION:

( http://su-etherpad.com/core-Jul7-BB )
Wednesday

7/7/2010

9am - 12pm

BB CL3 Brainstorm Research Foundation (Linda Avey)

Slides:

BB CL4 Life as a growth industry (Lori Giver)

Slides:

NT CL4 DNA Origami (Paul Rothemund)

Slides:

Suggested use:
Step 0: Check if someone else has already pasted your link.

If so, vote for it with a plus sign (+):

...

[12] AB

[13] CD ++

http://someone-else-beat-me-to-it.com

[14]

[15]

...
Step 1: Grab a number
...

[12] AB

[13] CD

http://someone-else-beat-me-to-it.com

[14] \x3cyour initials go here>

[15]

...
Step 2: Paste your link and (optional) excerpt and type an annotation:
...

[12] AB

[13] CD

http://someone-else-beat-me-to-it.com

[14] JG

http://www.slideshare.net/eekim/collaborating-via-artifacts

Slideshare presentation: \"Collaborating Via Artifacts:

Understanding The Real Difference

Between Online and Face-to-Face Collaboration\"

Great example of slidecasting and explation of how on-line collaboration works so effectively.

[15]

...
Step 3: Return to the original Etherpad ( http://su-etherpad.com/core-Jul7-BB ) and drop in your reference:
blah blah blah [14] blah blah
REFERENCES:

[0] \x3cadd your initials>

\x3cpaste your link here>

\x3cexcerpt and annotate>
[1]

[2]

[3]

[4]

[5]

[6]

[7]

[8]

[9]

[10]
[11]

[12]

[13]

[14]

[15]

[16]

[17]

[18]

[19]

[20]
[21]

[22]

[23]

[24]

[25]

[26]

[27]

[28]

[29]

[30]
[31]

[32]

[33]

[34]

[35]

[36]

[37]

[38]

[39]

[40]
[41]

[42]

[43]

[44]

[45]

[46]

[47]

[48]

[49]

[50]
[51]

[52]

[53]

[54]

[55]

[56]

[57]

[58]

[59]

[60]
[61]

[62]

[63]

[64]

[65]

[66]

[67]

[68]

[69]

[70]
"

http://su-etherpad.com/core-Jun28-AIR-CL1-Overview-AI-Robotics-Jacobstein

"Monday, June 28

AIR CL1 Overview of Artificial Intelligence & Robotics (Neil Jacobstein)
Neil Jacobstein

Slides: https://docs.google.com/a/singularityu.org/present/view?id=0AWIj9HPeinaFZGY2cDNjOXpfMzAwZHAzazY5OXM&hl=en

Audio:
Where are we now? Where are we going? How will we get there? How fast? Are all elements exponential? Lessons learned: early AI predictions and industry, Managing technical and business opportunities, Anticipating and managing risks..
9:04

Salim: Holland won at FIFA.

How many went to Luke's thing? [about half]

Will be quite a few of those over the summer, if you missed it.

Starting now with core lectures. Lectures in the morning. Workshops in the afternoon. Evening events.

Acknowledge unconference: fantastic.

How many for movie night? [most] Discussion after.

TP Survey was due - about half filled it out. Just initial sense. Please fill out.
One of missing students has arrived.

[hug]

SJP: Traveled from Australia

Sarah Jane Pell

Artist, commercial diver

Space analog projects

ISU alum

Worked on luna gaia 2006

TED Fellow

Working on art & science collaboration and intersection

Will be jet lagged. Look forward to meeting you all.
Dan will kick off with AI and Robotics. Welcome Dan Barry.
DB: Neil and I are co-chairs on this track.

One of really big issues: what does it mean to have robot overlords, I mean robot servants.

Introduce Neil.

Formal background

AI R&D 25 yrs for govt and corporates
Standford media program

AAAI for over a decade

Really proud to say Neil is a good friend.

Met here. At curriculum planning committee.

Weren't planning to work together, but he kept making insightful comments.

I'm thinking, \"how am I going to get to work with this guy\"

Low and behold he is chair for AI.

He was more interested in Network guys initially.

We've tried to bring AI & Robotics together.

Initial impression more than confirmed.

For all intelligence there, outstanding thing is his heart.

What really matters in this world.
Neil: Thanks Dan.

AI Overview

Nature has no departments, so we will be interdisciplinary.

Alan Turing 1950

Imitation game - now called \"Turing test\"

\"We may hope that machines will eventually compete with men in all purely
In 1950 Alan Turing published a paper about testing whether humnas can discriminate a sophisticated AI algorithm in which he introduced what was later to be known the Turing test: can machines mimick human intelligence in a way that is indistinguishable to a human)
Newell and Simon GPS - general problem solver

Herb Simon - 1957 predicted computer would beat human in chess in 10 years. 1997 it happened. That 30 year difference was significant.
Important to distinguish Narrow AI (specific problems) and Strong AI (the kind seen in the movie last night that involves the broad deep and subtle intelligence that we link to human intelligence)
[slide of tall contraption]

Shakey the robot -

http://en.wikipedia.org/wiki/Shakey_the_robot
Three development vectors:

1/ Analytical paths to AGI

2/ Reverse Engineering brain

3/ Application Paths to AGI

- two types

- AI - artificial intelligence

- IA - intelligence augmentation, douglas Englebart

\"Mother of all Demos\" overlapping windows, mouse, embedded video, outlining, hypertext

Not just showing off amazing things, Doug was concerned with augmenting human ability to deal with complex problems. Deeply concerned with amping up people's ability to face these problems. Brains are wired for local phenomena.

But we now have torrents of data and knowledge. Ancient brain architecture, formed LONG ago.
What's the problem? Why should we amplify our intelligence?

*Ancient brain architecture

*Evolutionary hardwired biases

*Limited Attention capacity

*Technology amplifies our power +/-

*Change exponentially accelerating

Obsolete government Institutions.Needs a complete overhaul to be able to solve our challenges.

Navigation requirements increasing rapidly

A lion roar - misinterpreted - resulted in death.We have selection for reaction and if we chose the wrong one we could die from that choice. Limited attention capability, channel capacity.
We have exponential change, governments need overhaul to deal with the speed of the modern world.

Much higher degree of navigation than we have ever had before.

Consequences of not getting this right are severe.
News isn't just bad because of our fear based biases for bad news.

We have some very severe grand challenge problems needing urgent attention.

What can AI bring to the table?

Represent the problem - in a way that machines can understand and apply.

Do inference, if this, then that, NOT that.

all we want to do is represent problems, control them in some way, and then draw inferences from that.
Expert systems - limited success. Why? Could outperform builders. Unfortunately, quite brittle. If knowledge was not in the system, the answers would fall off clifff. No graceful degradation. Difficult to maintain. Expensive to build. Very difficult to keep up with torrent of data from every day life. Acquisition, inference engine, but still -- a step in the right direction.

Range of systems like this to be shown next week.
Exponential curve of computer power has increased tremendously. Vastly more powerful than early AI computers.

1997: Computer beat Grand Champion Kasparov, fulfilling Simon's prediction.
Some observations:

The early predictions of AI were truly naive. Predictions 30+ years too early and had funding consequences. Customer expectation problems. Opposite of Tom Peters' customer expectation mangement advice: underpromise and overdeliver. People will not believe in it anymore.
In spite of this, exponential acceleration needs to be factored in. Now researchers may be unduly sceptical.

I am optimistic that we will achieve AGI, even superhuman intelligence. Be careful about making predictions as you'll be betting the project - find how you can add value at the beginning in the curve and then \"skate to where the puck will be\" - ride the acceration curve
Also interesting to reflect that tasks hard for humans turned out to be easy for early AIs. Irony, humor, empathy are difficult for computers.
Real world problems are domain and task specific. Solutions may require vast amounts of knowledge. We can utilize the power of massive data.
URLs - Uniform Resource Locators - point machines to documents.

Linked data enables people to find individual items in a document using URI

URI - Uniform Resource Identifiers - which a machine can understand and process
Untagged massive data now allows us to harness the \"unreasonable effectiveness\" of data.

IEEE Intelligent systems March/April 2009. Substitute billions of data points and non-parametric models to get machine translation.

Getting amazing results from very, very powerful harnessing of data.
[slide Nwanna's primary attribute dimension]

http://blog.chrisfournier.ca/wp-content/uploads/softwareagent_figure1.png
CALO - Cognitive Asistant that Learns and Organizes

http://en.wikipedia.org/wiki/CALO
There was the possibility for a system to do massive joins and create a \"lightning strike\" of data to the user
Embedding AI into robots: Sense, Plan, Act
Autonomous robots in cars for DARPA Challenge
Wifi connections of robots to play robosoccer
iRobot has developed robots for military - out of despiration, they produced a vacuum robot (Rumba)

http://en.wikipedia.org/wiki/Irobot
Open Source robotics - Willow Garage

ROS stands for both Robot Operating System and Robot Open Source

PR2 personal robot platform

Get AI embedded in physical devices in real world.

Also embed AI in built environment in intelligent rooms at: MIT, Standford, CMU, elsewhere.

Deliver some kind of augmentation to users around building and design.

NASA Hyperwall http://www.nas.nasa.gov/Groups/VisTech/hyperwall/

Display device - but not just a pretty display. Each panel has significant processing power behind it. Can be harnessed together to show patterns.
Second Life

Teleplace(?)

Build AIs into educational environments. Like this sphinx.

AGI:Artificial General Intelligence

To get to AGI, we have to do something a little different.

Project: OpenCog, Ben Goetzel.

Using Novamente architecture - go to site and contribute

How do we develop the components. Ben will be here next week, sharing his ideas.

Other threads available: image brain to reverse engineer it.

Understand principles

Electorencephalograph.

fMRI

- locate specific areas involved in specific kinds of processing

Understand construction of brain - diffusion tensor view of white matter

Diffusion tension gradients tell which way the white matter fibers are going.
Model of Auditory Pathway

http://www.lloydwatts.com/neuroscience.shtml
We also have models that were extracted from neuroscience research. Developed a chip that is able to discern speech signal from noise (Young 2003)

There has been a shift from brute force cmputing power to empowering standard PCs to do a very challenging task such as playing chess.
No doubt, we will reach AGI. Not just question of when. How we harness.

What happens after we achieve human level AGI. What values will we embed?
Next generation AI: image of computer based emulation of brain through reverse engineering.

[red line through computer brain]

We won't stop at brain's limitations. It is clear that we are going to break out of that prison.

Useful project would be to integrate all of human knowledge. People can master a couple of areas in a deep way, but no one is able to comprehend all of knowledge now.

This is a slide - every node is a different area of science. (http://www.plosone.org/article/info:doi/10.1371/journal.pone.0004803 ) This map is moving at exponential speed. Imagine the kind of velocity and accelleration of human knowledge.

AGI could be genuinely synthetic in integrating this knowledge
We will no doubt integrate AGI's with our own intelligence.

You might ask,

How?

What would it mean for you? your identity? your responsibilities?

Not only embed in environment, built environment, ourselves, but will control molecules and build the environment that is worthy of who we are.
Q: How much of AI technology ... you showed map. How much needs to built manually?
No hope of building that kind of system manually. Speed of knowledge increase will outstrip manual efforts.
Q: Is first step to build system to understand the meaning of stuff? Still years away?
We have Wikipedia now. Not doing inference. Systems that can. That is not a deep synthetic understanding.

Where do we begin? Get all data linked.

Huge corpus of knowledge.

Need systems which understand semantics.

Long way from systems which understand map.
Q: What is SIRI?
Massive joins of cloud services for specific tasks. Harnessing a number of different services. Does not comprehend huge amounts of knowledge to do that.
Q: Dmitry: You showed that we overpredicted. Are we getting better?

Always a lot of uncertainty. Now may underestimate what AI is achieving.

Try to get multiple data points. Look for counterfactuals. Always build in incremental advantages in your project so that you can deliver value at different stages of the project.
Q: We'll be better at prediction in 10 years!

Good prediction.
Q: David: Do we understand the 2-3 big breakthroughs that need to occur? Like DNA was a big breakthrough. From that framework ...
Key areas that will be pivotal:
Extract generalizations - find what is really common. Some can be done statistically. What are salient features? Will be very big deal.

More raw computing horsepower. You've seen Ray's predictions - we'll see based on raw power.

Other area: whole system for doing active semantics. Underlying meaning of words. So systems could read and learn directly. Would be huge.
Q: Do you mean by generalizations: This is a \"chair\" this is a rabbit\"?
Some systems can classify.

They don't tend to identify and understand general principles.

Bacon could recapitulate some Physics results.

But currently, this is a research frontier.
Q: Bryce/LA:

1/ To what extent is intelligence static versus functionally defined?

a property?

or a behavior?

If we see different behaviors, is it still intelligent.
2/ Tarsky's theories in AI?
On first question: both.

If a machine can play reasonable game of checkers, when Samuel achieved that in 1950s, the question became \"chess\" - then \"go\"- a line-drawing game. We keep redefining what intelligence means
Q: Maybe real reflection of what intelligence is. Perhaps intelligence can't be defined outside a context.
People keep changing the criteria.

We change because we are augmented.

What you can do with PDA or supercomputer is very different from what you can do alone.

Links to our evolutionary capabilities.
As far as Tarsky's theory - ecology of competition. One of many competiting.
Q: /Argentina: distinguish machine learning vs semantic based. To what extent do ...

I like AI because I thought it could be a mirror.

More I work on it, we will be able to be an AI, but it won't tell us much about us.
Depends on the way we build it.

If by reverse engineering, using same operating principles, it might tell us a lot.

Not subjectively, but a lot about human brain.
Other techniques, might tell very little about how brain works.
Q: Do you think this distinction between statistical and semantic approaches are really different?
Converging. Both can use probablility theory.
I don't think there is a never-ending war between those camps.
Q: David: In RK book, by 2045. Later, more whole humanity.

Basis?

What about software?
Graphs refer to computational equivalence.

Some rough correlation, but not same thing as intelligence.
Based on number of neural connections and operations per second.
Q: Does hardware progress faster than software?
Software will keep pace. Computer chess - at first was hardware. With Deep Fritz, software has provided huge value.

Predator/prey relationship. Software requires more hardware.

I've often waited for hardware ...
Q: 1/ List of conservative predictions - no breakthroughs, just growth of hardware and knowledge - where do we get?
Would challenge the idea that exponential growth alone will not produce breakthroughs, quantitative differences will yield qualitative changes.

In 10-20/25 years to AGI. Reason for error band, depends on how you define intelligence.

If you have specific idea, people might just work in that idea and achieve it early.

If you want an AGI solution, might take longer.
Q: More data -> AGI?
Not just more horsepower.

If collaboration around world, tremendous increases in processing power, just by incremental improvements, we could reach AGI, but it depends on what people work on.
Q: Blind spots. What can't we do? What will AGI never be able to do?
Very difficult to answer, given my blind spots.

We ought to run the experiment and find out.
Q: Sasha/Canada: Self awareness, creative expression. Part of AGI?
No areas of human endeavor are walled off from AGI.

No magic process (like creativity and self-expression) inside human being that is walled off from AGI.

We will build machines that do everything humans do and then some.
Q: Protectionism from humans?
Already seen it in low level form.

Seems to me that most important thing to do is to take care of the basic security and survival needs of fellow citizens. Address the Grand Challenges. Lower levels of Maslow's hierarchy. Let them live in a realtively high quality of environment and the push back will be a lot less.
Q: Manuel/Spain: Types of problems to solve are limited. NP
I think when problems are extremely hard theoretically, what we will see are various forms of attack. Machines will attack problems people will not be able to solve. Later, machines will push limits.

Dan Barry: You know that the problem is fundamentally difficult. You can come up with something close.

Instead of solving analytically, I'll solve numerically. Not going to solve it perfectly but can get close.
Q: Brazil: AI can lead us to AGI ..

lk

To clarify, I don't think incremental progress will necessarily lead to AGI. But it could, given exponential acceleraton.
Q: Narrow AI. Forecasting. Human - merge narrow AIs to simulated ... put in cloud. If this would be framework for general intelligence?
SIRI (recently bought by Apple), CALO

http://www.intomobile.com/2010/04/29/apple-acquires-siri-may-the-voice-powered-wars-begin/

Not final answer for AGI. Not satisfactory architecture. More brute force. We'll come up with more elegant solutions.
Q: Human intelligence - life experience. Building through nodes. What about crowd sourcing? Collecting through games, develop AI much much faster.

Protein structure in 2 weeks.
Really interesting.

Conflates issues.

Super human by networks of people

Really interesting and effective.

Multiple - achieves a kind of super intelligence.
Other issue: can we achieve an AGI by piecing together. May be possible.

Architecture for doing that is pretty week.

Not on wiki.

Maybe on opencog

Collaboration by adding tagged data and things can get you partway up the curve on AGI

Human-level AI and beyond will probably come through more elegant solutions
Q: Can we build a system that learns from people?
Right. David Canfield Smith built systems to watch, infer - what kind of problems humans are trying to solve. You can get interesting results. No one has taken it to the point of watching humans, then going off on own.
Powerful rule: use massive amounts of data. Will be good at classification, but not broad, deep features of human intelligence.

Interesting, but needs more architectures.
Eric/Italy. AI on human brain. What about other intelligence? Tropical forest. Studies look at this for collective intelligence.
People building on Gregory Bateson work. Feature extract properties.

Heterogenous fields that don't talk to each other.

Ecosystem is on fringe of AI. It would be wotrthwile exploring this.
Q: Shary/US What perspectives are active in AI community?

Computational?

Cog/Psy?
Cog/Psy - Simon and Newell. Beautiful book. Whole conferences on cognitive psychology and AI.

Less so with organizational behaviour.
People most needed now

Computational Neurobiologists - search to find really talented people.
Q: Eugiene/France: what lab to watch? What is most advance?

What could we learn about collaboration between robots?
Leading Labs?:

Depends on interest. Hierarchical Numenta, Redwood

MIT, Caltech, Stanford, CMU - All have some kind of AGI

DARPA community, typically one or more

Intelligence community.

http://www.numenta.com/

http://redwood.berkeley.edu/

http://www.csail.mit.edu/

http://robotics.caltech.edu/wiki/index.php/Main_Page

http://ai.stanford.edu/

http://www.cs.cmu.edu/~softagents/
Collaborative Robots:

Have to adapt to environment.

Quick feedback loop - plan/act/cycle

Fast learning from environment.
Interesting difference between static AIs.
Salim: Neil is core faculty. Time to ask questions.

Back in 7 minutes.
10:18
**********************************************************************************************************************
NOTE - we will continue with the EtherPad for Dan Barry's talk
Salim: NASA introductions
This is a lecture you will never forget for the rest of your lives.
AIR CL2 Adaptation & Cooperation (Dan Barry)

Slides:

Audio: https://docs.google.com/a/singularityu.org/leaf?id=0B0vcBiirD3NQOTczZTA2MzItMzVhNi00NjRhLWEwNTItYWUyYjc3MTQ0N2Mz&hl=en

AIR CL3 Robotic Communication (Dan Barry)

Slides:

Audio: https://docs.google.com/a/singularityu.org/leaf?id=0B0vcBiirD3NQZmI3NWIxODUtMGQyOC00ZTQ3LTg0YzktOTRiYjE2Yzc2ZWRk&hl=en
Dan: Neil has talked about AI. Now we will talk about AI in the real world.
That's not what I'm going to talk about first.

Principles of robotics.

I want you to remember

adaptation - respond to new situation

A second key word to take away today is cooperation trumps compeition in many situations.
Video - history.

German V2 blowing up.

Important to keep the pointy end forward \x3c\x3c key concept!

Now control system issue.

Really not good in rocketry.

No one was hurt in any of these videos.

$1 bn loss in 10 seconds.

Chinese - not just US and Germans that don't get it right.

Remember this is day one on the new job that you get this video.

Space X - just 3 years ago.

This one is pretty. Announcer is \"descriptive\"

\"It appears we have had an anolomly\"

Firestorm on ground.

You don not want to stand within a mile.

This is Cape Kenendy

More than 2 dozen cars are destroyed - windshields and tire rims melt.
So conclusion? Space Flight is Difficult.

You are getting ready to ride on these things.

People designing these rockets are bright people.

But from 1940s to last year, they blow up.

Every rocket, they put 4 in the ocean.

Why?

Why do you think when you build a rocket, using 50 year old principles, do they always blow up?

Everybody loses rockets?
Q: Complexity?
Sort of. Well understood. By the time Space X is launching.
Q: Explosive things on fire.
We do this with cars, trains and buses. We built a computer - see the wire in that sucker? It ran when we turned it on.
Q: Limits?

Q: Using compoents for first time?

Never tested in env in which it will operate.

They have to perform perfectly first time.
Why can't you test in env?

No zero g.

Test stand is different.

Talk to Glenn - pogo - a resonance. Sloshing of fluids. Doesn't happen unless suspended in air.

Very first performance is first time.

Second = no ability to adapt.

One regine. All they will do

No ability to recover - no intelligence. no adaptation. Fundamental reason they fail: Adaptation.
Difference between machines and biology is adaptation.

We adapt at every scale: Every time and place.

We have immediate adaption.

Tennis - intermediat forms - you learn over years.

Long term evolution

Now adapation is in our environment. Tranformed equation a bit to the point we are transforming the climate.

Climate change allows shortsleeves shorts on Mars.
Having said that, I want to show you form of adaptation the best I have ever seen.

Hope you can see this bit of algae. Holy crap - octopus. Think of adaptation. Watch in reverse.
Octopus is matching all the attributes of it environment.
Don't put octopus in a cage that has any openings at all.

[video of octopus sliding through tube]

This iswhat life is like with no bones.

Less than 10%of body size.

Seattle put octopus with sharks - and sharks died.

If in Caymans, don't dive too close to algae.

That is adapation in animals.
As a clinician,

rehabilitation physician
Extreme arthritus patient.

Husband has cardiac issues.

\"I can't go to store because I can't start car. Can't hold pen. Zip zippers.\"

\"We want to live alone - no assisted living.\"

I'm listening to this fiercely independent woman and thinking this is what I went to medical school to help.

She pulls out tube which has key in it.

Pulls out hook - big handle, button hook. Now you can dress yourself.

List of 8-10 very simple tools that transformed her from dependence to independence.

Trivial cost -> significant difference.

Simple engineering to transform people's lives.

Saw so many other people with similar stories.

Spinal cord injury patients.

Think of the fear. Won't move arms or legs for the rest of your life.

Even that very first moment, chance you will never recover.

It is everyone's nightmare. First thing people want to do is commit suicide.

They ask physicians to let them die.

They have reactive depression for a long time.

Physician's role is to realize it is a panic attack.
Turns out, 3 weeks out. Meet others living indpendent lives.

You can live productive lfe.

Make friends.

Mind makes decision how to push past that.

We literally duct tape hands to oars. Get them paddling on the river.

Back to minimally assisted liveing.

All throug adaptatoin.
Something that happened to me: Not a disability. Positive adapation.

Our bodies change so fast, they adapt to zero G.

you can get lost by being upside down.

When we train, read book, flip switch.
Does space make people stupid? Why going 1 2 4, skipping 3.
Q: Disorientation?
Negative training.

In space, you are floating. You hand is floating. So your hand and the list float.

Your eyes land in a differnet place.
Now, mark through steps.

Takes 3x as long in space due to external adapations.
Best story. Pilot in right seat. Only person after engines shut down for doing important stuff.

Hydraulics working. In right orbit.

Buddy says, \"I was trying to be a realy good pilot. Everyon having a great time. One guy has his hand in front of my face. I got annoyed, hit the hand. It is his OWN hand.\"

Lost corpre-hensive? sense.

Used hands to find where hands are.
At bedtime in space, My own hand was banging into my head. Only happened that first night.

Don't know whether adaptation was to sleep through bumps or stop doing them.
Rookie's job to make patch.

by the time you are on your third patch, you just want you name spelled right.
Supersonic jet.

Houston to FL, hour and a half.

Talking in cockpit

T38

3 min in Shuttle.

LA to Boston in 10 min on Shuttle - if you are late to work or something.
Talking, going to go tonight? Finally time. The dream starts to come true.

Crank the sound

Shake and Bake

[Shuttle liftoff video]
2 min in, drop big boosters.

Only 3 Gs

8 1/2 min to orbit

Parallel parking at 5 miles per second

ISS at 17,000 mph

\"If not careful, you're going to hit that thing\"

Opening hatch. Bringing home crew - we were two months late.

Commander was REALLY happy to see us.

At this point, set record 10 people in space.

Safety briefing - what big red buttons not to push
Work you do is technician level work.

Batteries

Radio repairs

Installing mufflers - fans were loud

Fixing squeeky doors.

Trained on earth - 3 at a time. In space - 1 at a time.

How would you practice on Earth?
Culmination of Astronaut career is going \"outdoors\"

Spend a day preparing

You just learn how to do it while you are there.

Takes 2 hours to put suit on

You really don't want to miss a step here.

Marine colonel - renowned for following steps.

you HAVE to check off each step.

Opening door, stepping outside.

I'll talk later about beauty of environment.

There is a robot there. Very little autonomous robitics.

Always something goes wrong.

Bolt doesn't drive wiht power tool.

Ingenuity. Fix something right now.
Back indoors, work out at gym. All spring driven. You lose bone and mass in space.

Bungee holds you on treadmill.

Stupid astronaut tricks. Spinning people \"My eye balls are going to fall out!\"

Put students in middle of classroom - that's where they cant's get away.
Playing with food.

Tons of switches. Dog puller(?)

Beautiful , silver winged dragonfly.
Mr Junior Science - magnets. Gave me behavior I had not anticipated.

No convection of fluids.

You can drink out of ball. Won't stay in cup.

Tammy on 5th flight. Goldfish. Loot at dynamics of water ball. Cracker Goldfish (no animals were hurt).

Rest of crew came to help.

No convection - study surface tension in isolated environment.

Used dental floss to bring two fluid balls together.
We do a lot of student experiments.

Disco Shine. Stayed up about 6 months.
45 minute task coming home. Flashes are giant fireballs. Like paparzzi.

St. Elmo's fire.

Shuttle is like slamming into waves on a boat. Rattling and shaking.
It is aglider coming in 8,000 miles away. Drops like rock. You've been in sapce.
No g's for 10 days. You do it. Training takes over.

No gear until 300 feet off the ground.

done these - and everyone has worked.
Roll down runway, let go of the shute.

\"Wheel stop.\"

You've done this amazing thing. We all went and did this thing together.

You body says, \"Im not done with you\"

You are in your chair, exhausted. Ready to be greeted by friends and you can't stand up.

Legs go \"nope.\"

You sit there.

I know in 9 days my muscles are not the probelm.

Why can't I stand up?
First, a story. When you first get to space. The pilot and all. Woohoo, I'm in space.

Wife said, \"Hey that's Dan\"

What am I going to do: Tom Hanks. Floated glove in front of his face.

\"I really am in space\" unbuckled.

Dreaming of flying.

Push off from wall. Seeing other wall. Realizing I am going to hit it.

Half way through the cabin, I start kicking.

Vetrans in corners - so rookies don't kick us.

Tunnel - push off with pinkie.

You fly down tunnel. You see yourself start to drop.

Your body is tilting, not translating. so you kick with your feet and cram your head right into bulkhead.

Within 3 days - you konw your center of mass and fly like superman.

Next flight - home, not house.

People there 4 months have ADAPTED, learned how to live in space - dancing.
We are sitting together - not just physical adaption but also a not just physical adaption but also a mental adaption. Tasks done.

\"Lunch time\" we don't do that in space.

Staion guys do it.

Working in space, living in space - you can't work all the time.

Human beings need to have some downtime.

So - back to the issue - what isoing on with my legs.
SJP: Body memory.
Legs would push you through ceiling if used!

Bed pushed so hard. Felt I would blast through ceiling.
They say, \"Great job\"

Technician hands you clothes. You leave them right there. [holds hands out in air]

And what do they do? [smacks podium]

Gravity sucks.
So they tell spouses of returning astronauts: \"Don't hand them the baby.\"

[applause]
It takes a few days - no driving car. Couldn't fly plane.

Could not walk with eyes closed.

2 year later, 3 years later flew again.

No conscious difference - took 1 day.

Next time - 2 hours to readapt.

How can you take 3 days, then years later do again, body had figured out how to do that for you.

Nobody knows how that adapation occurs.
SO, what does this have to do with robots. Supposed to be a robot lecture, right?

Robots are not very adaptable.

Things that you think of when you think of human intelligence

are wrapped around this idea of adaptation.
YOu can divide robots into autonomous and non autonomous

General purpose (a rumor) (no particular function but able to do just about anything) vs. special purpose (in a factory)

therefore 4 possible types based on the combinations of those two types
EXAMPLES:

I. Autonomous special purpose

Video: Autonomous welding robots, people on floor is in very serious danger. No joysticks

Video: Warehouse robots: Kiva (http://www.kivasystems.com/index.html) a company that creates robots to carry items in a warehouse. Autonomous and special purpose. A huma in the loop, robots bring material to the humans. , Andy Barry use to work for Kiva

Video: Robotic fish

Video: Flowers combining robotics and art
II. non-autonomous robots special purpose

video: articulated hand controlled by a real hand, robo astronaut, robotic surgeon (da vinci)

video: big dog! walking robot for carrying heavy loads, Boston Dynamics mix between autonomous and non-autonomous

video: petman, two legged version of big dog

video: packbot, goes up stairs, durable enough to go through windows and jump off cliffs and water proof

\"better than a timex watch\"
III. Autonomous general purpose robots

video: leonardo - gizmo looking robot, expressing emptions towards elmo vs. cookie monster

video: willow garage opens door

http://www.willowgarage.com/

video: swarm, 278 robots communicating with each other, co-op behavior among many robots

[?]
When do robots NOT work:

example of robots and humans working together, some times it's good to have humans other times robots should stay in control

example: video of famous captain Sullenberger landing the passenger airplane in the hudson

Plane hits a flock of birds. Engines catch fire.

\"Lost thrust in both engines\"

Now heading to water. Very smart. If headed back to Laguardia, would have had crash.

Negotiation for Tetterboro landing. Pilot still hasn't turned. Focused on landing safely in Hudson.
\"Newark airport in 7 miles ...\"

Still trying to restart engines.

Out of ideas.

Lands in water.

That's it. Everybody lived. Everybody survived that because you had a person on board who understood the priorities. Person sitting next to him. Solved problem in real time.

Guy on normal takeoff. 10,000 hours. Notice how quiet. ONE comment on how pretty Hudson is. Cross check.

Captain recognizes hit. Takes control of the aircraft. Calls for procedure, rather than trust memory. He has already done boldface. Get started before you get book out.

Go through and cross check each other.

Cockpit warnings going on.

Trying to decide.

Methodically - assessing state of engines.

Air trafic control is helping.

Started to return to field. Then you hear him say, \"We have the runway ready\"

He didn't KNOW he could get back.

People on ground would die.

Crash in Hudson - no one on ground would die.

Other thing, over Hudson - has a shot at having some survivors.

Trying to get both engines. Guy looks out and sees Teeterboro.

OK, fine.

He then makes assessment, that he might, but might not. Waits to make turn.

If he gets decision YES, then he can commit.

Hoped engine relit.

Not going to happen.

At some point he puts focus on landing in river.

Flaps down.

Most brilliant. He says, \"Got any ideas?

He is ready an dable to listen for someone to tell him ...

that is having real control

think of all the variables, controls.

Thank the lord there wasn't a robot.
Contrast Colgate (?) chat about things.

First officer see ice. \"Very uncomfortable flying in ice\"

Never seen this much ice - 5 min before plane goes down.

Complacent.

Robot part of plane starts to shake the stick.

First officer raised flaps, made plane unrecoverable.

Two things - crew on top of it, other crew didn't take adavantage.

Not always true that humans are better at responding/adapting.
Given what you now know, I come back to opening ceremony question:

What good are robots?
Only Shary got back to me. Kudos.

So, it remains an open question to you.
I will tell you what robots do in my house and my company.
Nothing I can't do better myself.

We are cheering that it can fold towels.

Can get pizza and that is about it. Learned a few things, people are honest and never stolen from the robots and adapt quickly after working with the robot.
The point is to transport a person from dependence to independence using robots. Add the element of intelligence to the robot and it transforms the persons life.
HOMEWORK: What should robots do in my house? How do we make Q/A time more interactive with everyone?
12:00 Lunch
[end of presentation]
"

http://su-etherpad.com/core-Jun29-NCS-CL1

"Tuesday

6/29/2010

9-10 am

Brad Templeton/Salim Ismail

NCS CL1 Network and Information

Slides:

Audio: https://docs.google.com/a/singularityu.org/leaf?id=0B0vcBiirD3NQYWQ1YTQyZTktNTNkMy00NjBlLWE4OGEtZTJkYTFjYmJlMGRm&hl=en
Presentation slides: http://www.templetons.com/slides/singu-intro.ppt

All Our Ideas crowdsourced questions: http://www.allourideas.org/gsp10ncscl1
[Networking for Dummies Books: http://www.dummies.com/store/Computers-Internet/Networking.html]
\x3cnotes start here>
Color card definitions

Green - Agree

Red - Disagree

Yellow - Slow down (content) / I don't understand

Grey - Speed up we know this stuff / Move on to another topic
hands moving apart (sideways) - speak slower, speak more clearly
Computer's are getting faster but the bootup speed remains the same
What makes a revolution?

- Build technology for early adopters

-Nobody who can tell you \"no\"

- early adopters are usualy stupid people with too much money :) - willing to waste their money to stay ahead of the curve

-early adopters provide a market for new tech

-Insatiable market demand for doubling, less demand for incremental improvement

-Disruption!

no permitting process

The personal computer had its explosion in many ways as reaction to restricted mainframes.

People bought their own computer and it was THEIRS.

It is amazing how having to ask permission slows down development.

You want an open and hackable platform, you want a culture of talent. Competition to get to these early adopters.
Store full of stuff, including wire-wrap boards. For individuals to be buying this stuff for $20 drives revolutions.

Market wants things 2x better, not 10% better.

Thanks to slow boot - you want and need a faster computer.

Your old computer is seen as a dinosaur.

Moore's Law is not a principle of nature - it is a marketing order.

It became self-fullfilling

$1 bn was spent making the fabs to make those chips.
Book referenced: http://www.amazon.com/Free-Future-Radical-Chris-Anderson/dp/1401322905
Revolution in Networking - the internet broke a lot of rules.

Packet-switching networks were \"smart\"

Big institutional approach.

Internet tried something much more basic. David Isenberg [?]

Just get packets, put an address on it. Network does not try to be \"smart\"

Instead it only promises it will do its best. Not a guaranteed system.
Eric: Can you explain what a packet is?
On phone, you had circuit.

We may be meeting Vint Cerf.

Break into a few hundred bytes of data. Mine and yours stream together and get split later.

Some to Boston, some to Santiago.

Idea was, very simple system. Packet switching.

Let people innovate in boxes at endpoints.

That lets people do the innovation where no one can tell them no.

When someone comes up with something cool, they can do it.

Actually cheaper to fix the capacity and congestion problem by buying more bandwidth, rather than trying to get smarter.

Break network neutrality rule ... may not be best way to go. Misguided.
Another important thing - pricing related.

Best invention, not packets, but cost contract.

I pay for my line to the middle and don't care what goes in between.

Each building wire to the center.

No body could say no.

First famous application - coffee cup or fish tank in Internet lab.

Everyone wanted to see this.

If they had been charging per megabyte, guy would have had big bill.

Basically it let people experiment.

YouTube - people experimented.

No \"bean counters\" (accountants)
In networks,

The internet isn't free but because of the flat pricing, it appears to be
Optical fiber now at 69 terabits per fiber with Dense Wave Division Multiplexing (DWDM) - equivalent to everyone on earth being on the phone at the same.

This includes multiple colors of light so you can send more information at the same time
Free Space Optics (FSO) - 100s of gb/s, terahertz
People think there is a limit to the amount of spectrum available.
This is the actual of Wi-Fi miracle.

Take piece of spectrum. No body thought that

One very interesting rule. Your problem if you got interference.

I think that resulted in the greatest innovation in radio -

cordless phones

sprung up out of nowhere.

This lesson is slowing showing up.

Cognitive radios - not blind. Understand which bands to use based on which have the least ammount of interference.

Whitespace - frequency allocated to broadcasting service but not used locally, FCC ruled that these can now be used if they don't interefere with assigned broadcasts.

Lot of TV channels 2 to 59. Only 11% still get TV over the air.

Not 60 broadcasters. A few dozen at most.

Smart device could see no one is USING channels, jumps in and takes advantage of it.

This is being fought.
As long as big companies don't get in the way.

Glass roots revolution.

People buying fibre connecting to next block.

Freespace links. Changing the world.
Not a lot on programming languages.

In the beginning, low level - close to the machine.

Early on, probably C.

More common now is higher level - further from the computer. More abstract.

[JG: I like Python for example]

Result of this scripting is that programs are bigger and slower. Rely on features like garbage collection - reallocating memory that is no longer being used.

Different mission.

Waste resources - memory, hard disk,

and not waste your own time.

Sooooo many transistors, no need to make layout perfect.

Human time became more important than other resources.

So software bloated.

Fundamental theorem: \"Every problem can be solved by adding another level of indirection.\"

You have a name for a name that points on.

Let's you change things - implementation - at any level. Very powerful.

In virtualization, there is a lot going on between intent and implementation.
Next presentation will cover this.
PLE lecture. Another trend: software recall.

Legal consequences.

Companies can update software remotely.

Each package starts a new daemon. \"Is there a new program ... is there a new program\" Hence slow boots.
PlayStation 3 - disabled capabilities

IPhone

Amazaon 1984 \"unbooked\" deleted from Kindles.

Could they have picked a better book [?]
People are trying to abstract even more.

Virtualization - something that looks like a computer inside the computer.

You might have Parallels - looks to Windows like a Windows PC (running on Mac)
Build a room that puts virtual computers inside computers.

Computing on demand.

We can sell you a hour's time - Windows Linux (not so much Mac).

You can buy 1 or 1000 for an hour.

Trend: Cloud Computing.
I will rail about second one - take personal data and store it on machines owned by 3rd parties.
This is the trend. Can delivery massively parallel computing. 1000s and 1000s of computers on demand.
Goes from big computers (main frames) owned, to time sharing, to PC, to client-server mix, to finally this cloud thing, \"Web mosft

something\"

People trying to do more in servers.

All desktops - all that matters is running Javascript.
Media.

History has axis - serial vs browsed. Push vs sampled.

E-mail 1, then another.

Blog, serial.

Magazine.

On-line is like that.
Browser - hop around. Not prepared sequence.
Two types of serial -

IM

SMS

vs

Serial you ask for:

Blogs

E-mail
Sampled is old/new.

Twitter. Interesting from other streams, you don't follow whole feed.

Other than surfing TV, this is a new type of media.
Other axis: reader vs writer friendly.

NY Times is reader friendly. Pyramid style.

vs

Wiki is writer friendly - easy to go in and change. Hard to follow.

Browse, but otherwise very writer friendly.
Use categories to help understand.
All new media create information overload.

FIND new? Ah .... I don't get enough! Too much signal.

Signal to noise - how much good to useless.

Too much good information on Internet.

Problem is not finding but filtering - you will deliberately NOT read all that is good.
FREE model. David Rose.

When you want ad revenues, efficiency issue. How many ads did I show to people.

How many leads?

TV ads - 30 seconds, gets 2 cents to network to show that ad.

You get hooked to buy product.

Google gets 2 cents. How many spend 30 seconds. Eyes spend 1/2 second.

google is deliveering same value to the advertiser in much less time.

Delivers value of ads to you 50x better.
Coming up: next 10 years of hardware and software.
Open source Wed. Visit Google

Quantum computing.

End of the week: social media/networking.
Wed: City of the future - wired and full of robots. Delivery robots take things around.
www.Robocars.net (brad's website - can see his current book project)
Workshops on the internt of things.

Room full of cubes in silicon valley - not good site visits.
Visit Frye's

Exploratorium.

Tech museum

Hacker Dojo [JG: Yes. This was cool!]
At NASA, Homebrew robotics at building 23.

Kathryn has calendar here: []
Jan Gray.

Went off to work at Microsoft to examine strategy.

We built them one way, now multicore.

Changes what computers, especially software, have to do.

Will get much more technical.

Moore's law is at the basis of everything.
Jan:

- Scaling Up Computing Power

Slides at https://docs.google.com/a/singularityu.org/leaf?id=0B6vjqTaW5opaYWU0MWJlODctZWQzZi00MTk1LThiN2UtZDlkODNjMWMwODAw&sort=name&layout=list&pid=0B2Ij9HPeinaFODJjNjVlYTItNTlhMC00YzI0LWFlZDEtM2U3NTg3ODc5OTNj&cindex=1

Audio: http://bit.ly/9ORmN4

I was born when Moore's Law was framed. Hope it will keep scaling up.

Around 2004, ebb in performance scaling. Run from 1970 to 2000. Doubling every 2 years.
1 MHz to 3000.

Each clock cycle did more.

Heat disappated went up.

Rounding off of clock speed: 3 GHz

Yet we've gone from 10,000s to millions of transistors on a die.
do we have a good c?

architectures - how to spend them?

How does software change?
Some electronics. MOSFET

Transistor is three terminal device. Source, drain. Control gate.

Channel has semi-conductor material.
If NO voltage, channel doesn't conduct.

If positive voltage, you get conduction from source to drain.
[Video - sand to silicon chip]

Make one large silicon crystal.

Slice it.

Photolithography.

A resist. Expose through mask with many many copies on round wafer.

Etch, deposit.

Build parts of transistor - 9 layers of metal.

100s of saws cut it up.
Several hundred billion transistors on that one die.
[Chart from S. Borkar]

Each 2 years, we go into a new technology Node. Smaller by 0.7x. Doubling in 2 year.

Quadrupling of transistors on a die.
Quarter of a trillion transistors on a die by 2018.

Rate of speed improvement is going down. Not as well as they used to.

Then it will slow down to every three years after that
Extend 193nm refractive lithography

Optical proximity correction, double patterning
13nm EUV lithography

soft x-rays, vacuum chambers
Transistors:

new material in gate, insulator, channel
DRAM - capacitor, 2x/3y

FLASH PROM - floating gate 2x/1.5y 64gb at 25nm
3d: stack cells vertically

resistance change/phase change RAM

amorphous/cystalline
Slowing of transistor doublings (now we are 2x in 3yr)

But.. cost halvings still continuing - \"22nm ought to be enough for anybody\"
Computer Architecture

1986-2002, computers 2x in speed every year but now its leveling off
Instruction Level Parallelism (ILP) [cut and paste from slides]

Spending Transistors on Performance - run programs 2x as fast
L1 cache reduces memory access time

Branch prediction for determining where the program is heading

Vector operations FPU's for working with videos/graphics
Over 30 years - CPU cycle time 1000ns -> 0.3ns but DRAM access time 500ns -> 100ns

sometimes progress stops for hundreds of nanoseconds
\"The Power Wall\"

Over last 30 years: die size larger, voltage 15->1v, frequency 1->300MHz, Power 1->100W

like the heat of a rocket at 1000x and surface of the sun at 10000x, just can't go there
P porportional to CV^2f
the power wall is the limiting design constraint
\"Complexity Wall\"

diminishing returns in wide-issue machines

only a tiny part of the die is actually doing processes
\"Explicit Parallelism - Multi-Core\"

much more area efficient

power wall is still there, how to solve?

finesse power with lower V, lower frequency, sleeping of transistors

finesse memory wall with memory parallelism

can't think of anything better so let's see if we can get the programers to get it to work

32 cores = 1 trillion floating points per second
Single-Chip Cloud - 48 individual cores
\"Scaling Up Memory Bandwidth\"

TFLOPS need TB/s of data.... 100s of GB/s not nearly enough

solution: bumping, die stacking

\"Silicon Photonics\"

should surpass copper traces for non-local interconnect

optical links at TB speed

Die stacking + Photonics = adequate bandwidth for next generation
\"Graphics Processor Computing\"

100s of cores

1000s of threads

Graphics module is very accessible to scientific and technical computing.

GPU use is still \"black magic\"

Game enthusiasts buy $500 boards and drive supercomputing.

Nvidia chip with 500 cores.

http://www.nvidia.com/content/global/global.php

http://www.nvidia.com/object/cool_stuff.html
Systems on Chip, Network on Chip, Modularity, Heterogeneity

database with large cores, small cores, media cores etc... mix and match for different problem domains

After 2015, a lot more fault detection on chips

Eventually to the point that computers are modular/stackable

iPhone G4, optimized for low energy
Eurocloud - research project that makes 3D cloud computing on a chip
Break at 10:26

Brad: I hope this was good. [Green and red card votes.]

Building a chip today is like trying to paint art with a paint roller
Connor: SU exponential technologies. Age of abundance.

How will computer design / chip /datacenter design when energy is limitless.
Jan: All large scale production systems will be expensive, hard to deploy. Don't know where ..

IF we had unlimited energy, constraints would go away until planet melts down.
Brad: Unless we come up with technologies, like reversible computing.

They get as hot as surface of the sun.
Jan: We'll just use lasers to put energy in space.

Brad: Or put computing in space. [Dmitry?]
Eric: Glass revolution. Explain a bit more. People-to-people.

Brad: In English, Grass roots is something coming from bottom.

Markets have exponential growth. If networking is done at grass roots, connecting neighbors with Gigabit.

Like Wi-Fi - always a hotspot somewhere near you.

Sometimes give away free.

Revolution of access.

Take that model so you can go to local electronics store.

Fibre is safe.

Not worth stealing - unlike valuable copper.

Sends immense amount of data that does not involve phone companies.

[applause]
: Wondering about cloud. What is next?

Brad: Stratosphere. Proposal to put computers in space. Actually latency to LEO is quite good. As good as using something on your desk.

Other proposals - quantum computing next week.

Strange rules of quantum physics to do computing in small groups of atoms.

May still have clouds of quantum computers.

That will stick around.

Jan: 50x or 100x cost reduction of processors in the cloud.
Brad: Buy computer today, run for 3 years. What is most expensive part:

The ELECTRICITY.

Not fixing that well right now.

Home energy track may address this.
Dmitry: Vacuum of space? Cold? Problems.

Brad: Only cool by radiation. Not so good.

Dan Berry- when they take laptops into space, they must put fans on them because cooling does not happen by convection in AIR.
: So the cloud is built and operated by companies at the moment. Should it be?

Brad: Next session.
Brad (GSP Brad):

Jan: PhysX - software physics interface. Discontinued ?agea?

Now in GPU. Gathered they felt scaling of GPU did Physics stuff without using 2 chips.
Sarah/ Australia: Potentials for supercomputing involving biological systems? Interfaces?

Jan: Lovely image of chip + neuron.

Inspired architectures - neocortex type in hardware, very interesting.
Brad: DNA computing?

Very slow, but for free, DNA will grow and may come out with calculation.

Jan: Biological circuits. DNA has nice property of complementary base pairs. Some parts of the strand bind tightly. Others left dangling. You can do AND gates. Strand activates whole other computation.

Not aiming for TFLOPS. Rather simple computation to unlock drug at particular site.

Kind of taking experience with computers and getting state, amplification in DNA strands.

Sky is the limit on what you could do with this.
Brad: You guys have been very interactive.

Tony / Korea: End of Moore's Law? Society is depending on it - some industries would close? Implications of plateau?

Jan: Every technology has one of these cycles. Fortunate technologies go through phases, slow down, turn to new technology.

Theme of this talk - uni-processor slowed, however with more cores - with parallel software we should get back to fast rate of scaling.

Not stopped growing - still getting 10-15% better. That is amazing.

We will either transition to something else or cheaper ...

Brad: Pause now. Come back in 7-8 minutes.

10:43
NEW PAD AFTER BREAK:

http://su-etherpad.com/core-Jun29-NCS-CL3
[end of presentation]

All Our Ideas crowdsourced questions: http://www.allourideas.org/gsp10ncscl1

Please take 2 minutes to vote for suggestions.

"

http://su-etherpad.com/core-Jul19-SPS

"7/19/2010

Monday

9 am - 12 pm
OUTDOORS near Building 20
SPS Session: Implications of Galactic Communities

Dan Barry
How fast do you need to go?

If you are pushed at 1 G throughout trip, you can get 6 nines.

99.9999% of speed of light.

What's going to happen, we are going slow.

Earth time and ship time are the same.

As we start going really fast, Earth time slows down.

It allows you to go to really far, distant places.
Explanation of why times slows down on a ship.

Dilation.

What is the deal with relativity.

Speed of light looks the same, regardless of frame of reference.

Walking along together - he and I are not moving relative to one another, but we are both moving relative to you.
Speed of light is going at the same speed regardless of perspective.

Why does light always look the same?

If it didn't then you could go the same speed as light.

And, if you did that, you could look at a light beam.

This would violate physics.

It is an oscillating field.

Einstein said, you can never catch one.
Sounds like a cop out.

Other people came up with other explanations.
But you base predictions on your cop out answer.

It might have been Dobachevsky.

But Einstein got the right answer.

Trick is when your hypothesis predicts physical things,

people do those experiments.

Newton was on top for awhile.

Someday someone may replace Einstein in terms of accuracy.
Clocks go tic-toc.

Now what happens if I move the whole thing.

Does it change the rate of tic-toc if you move it?
It changes the distance the photon has to travel.
You can take that to the bank.

It all boils down to pythagoras.

A lot of you look skeptical.

You all now understand the entire basis of special relativity.
Inside your ship, time is running slower.

Alpha Centuri - 4.3 light years away.

You cannot get there any faster than that.

If you get on ship accelerating at 0.1 G - it will take you 13 years to get there.

12.x on your calendar.
I see you get there in ~6 years at 1 G. You get there in 3 yrs.

You can get to A C in \"2 years\" if you accelerate at 2 G for the entire trip.

Spend time -

Who wants to go to Cyrius - 9 light years away

0.1 20 17 on board

2 G? 3 yrs on board
Bettle 520 l y away - - 7 yrs at 2 G

Center of galaxy - 10 lightyears
Andromeda - 15 yrs.

On Earth, 2.2 m years.
Round trip to Cyrius - 18 years on Earth. You'll be 6 years older.
Q: ???

Got to figure out where it is going to be when you get there.

Orbital mechanics.
How fast can you ultimately go?

What is the limit?

Energy.

Can only slow down so fast. True.

Mass. Another effect - as you go faster, you become more massive.

You also become shorter in the direction you are travelling.

Ultimately, as you compress these things, you turn into little black holes.

So, that occurs,

let me just show you

black hole has characteristic mass

Equation is very simple

Radius = 2 x gravitational constant / c^2 = 1.5 x 10^-27 m/kg

A very small number.

For 1 kg, say, you would have to shrink it to 10^-27 m in radius to turn it into a black hole.

Pop.

It becomes a black hole.

Electrons have radius 10^-24 m.

So what mass would electron have to have?

Mass would have to be 10^3.

If you can make your electrons be 1000 kgs, about a ton.

If an electron weighs a ton ... well, that is hard to do.

But it will when you get up to 16 nines.

99.99999999999999% of the speed of light.

When you get going fast enough,
Maybe your protons go first.

I didn't do that calculation.

Just as bad for you.
Erez: Is that a relative effect? So outside they think you are a black hole, but on board you are not?

Yes.
So, what does this mean.

Kids can go to college on Vega.
It means, I can order a pizza from 10 light years away.

Dominos makes the pizza with 12 9's. Pizza can get to me in an hour, still hot, but 10 years since I placed the order.

Guy gets paid 20 years later.

So, what does that mean? A stream of pizzas, constant - you need to trust the people you are dealing with when delays get long.

10 teams.

Break up.
"

http://su-etherpad.com/core-Jun30-FEE

"

.Wednesday

6/30/2010

11:30

David Rose
Future of the firm
Slides: [Jonathan Badal]

All Our Ideas: http://www.allourideas.org/gsp10fee29jun
Kathryn will review later looking for any [ref] markers--She and TFs will fill these in with references, but all are welcome (encouraged) to add references / put in as many details as are available.
11:38

David Rose: Making connections.

Going from genetic sequencing.

Romer last night, putting a billion people in Australia.

Economic aspects of business and economies.

How robotics and stuff effects real world and gets to market.
I will try to speak slightly faster than I usually do so you can get lunch.
Lunch in upstate New York.

Crystal Bay. Near marina. Not built recently. Sitting in large industrial park.

Why?

Found picture showing same place with a whole lot of other buildings.

What was this?

Turned out it was once the world's largest yeast factory.

In the lifetime of my parents, was the biggest in the world.

Self-contained city. Own docks.

What is it that created this enormous company?

That made it disappear so only left is a restaurant.

Technological changes:

Outsourcing

Crowdsourcing ...

Offshoring

New way of business: Mass Collaboration.

Very different from 20th century.

Wikinomics.
What is the future of the company/firm.

Go backwards to see why they exist in the first place.
Adam Smith. Father of economic capitalism.

The invisible hand.

People acting on their own, guided by free market, will get things done.

If people do what they do:

fix

build

etc.

Then they do deals, would not need companies at all.

What changes that?

Why did companies get started?

Ronald Coase: The Nature of the Firm 1937 [ref]

Something has to get in there and futz (manipulate) with it.

That thing/person is the entrepreneur.

What are all the possible reasons to override freemarket.
1/ Some people like to work for someone else? No. Sub-optimize.

2/ Some people like being bosses.

3/ No one in business is sub-optimizing, it is consumers who want to pay more. [BRANDS?]
As technology specializes and scale and distance increaase complexity increases requiring more _____
If you are using a lot of people to do something, more-and-more people involved.

Can't just grow and sell rose.

You get transactions from people working together.

What are transactions?

What do you charge.

Make contracts.

Fixed price - what about risk, who takes that.

How are profits divided up. Things change - then what.
Catch to each is \"Cost\"

Every time 2 people get together - transaction costs.

As long as it is cheaper to do things yourself than do transactions, you will do it in house.

And your house will get bigger.

Ford - one thing in garage.

Another with factory.

you will not buy one tire each time you built one car.

You may want tire plant in your operation [vertical integration]

Therefore you will get larger and larger and larger
For that, Cose won Nobel Prize in 1991.

This is why firms grow.

Guy still at U of Chicago.
So Ford.

Scale up. Requires lots of components.

This is where it ends up. Rouge River Plant - one honking big plant.

100,000 people working here.

90 miles of its own railroad tracks and 30 miles of its own Roads

[JG: This was in my backyard growing up and let me tell you -- it STINKS - and is mostly a wasteland now]

Found out that paper was big external cost turned out transaction costs were cheaper to make their own paper to print their manuals to put into their cars
[Slide: IBM \"city\"]

1000s working there. Now gone.

Why?

Because, over half a century, it became cheaper to go back out.

Outsourcing.

When you get source and you have to go through large bureaucracy to do one thing you lose economy of scale, becomes better to outsource
Externalities in effect

A/ Cost

B/ Can do things better outside. Find innovation.
Angel investors rarely see IPO (Initial Public Offering). More often acquired = outsourcing of innovation.
You can outsource virtually anything these days.

IT

Human Resources

Facility - lease back

Manage real estate

Accountants

Tech support

Mfg

Overseas shops

Outsource your community management - on Twitter for you in India.

Outsourcing market now an enourmous number.
Example: local Conde Nast titles on news stand.

Garden Design magazine

One person produces entire magazine.

Outsources photos, writing, subscriptions, etc.

Now replaces what took hundreds before.
Example of a crowd source company:

http://www.pond5.com
Cable channels use this. Think of Space Aliens. Used to be like a film production.

Pond5 has 300,000 clips. You can upload, obtain picture, say, of Hangar 1

at NASA.

http://www.nasa.gov/centers/ames/home/2008/hangar_index.html

Put on a voice over and - without ever having a film crew or even a camera - a whole cable show.
Florists

National retail associations (FTD = Florists' Transworld Delivery?)

FTD clearance system

National phone retailers (1-800-Flowers)

National web retailers (ProFlowers.com)

Individual flower farms (farmstogo.com)
http://en.wikipedia.org/wiki/Florists'_Transworld_Delivery
Columbian florist will ship direct to you.

Using FedEx.

Decrease in external tranaction costs.

What is interesting is global scale.
Every company must think globally. You can not think locally.

You have to assume your business should be global.
Angelsoft.

Started with a global concept.

We are - today - this teeny-weeny company, leads the world in angel technology.

Spain coming.

World business association

50+

Suppliers can be anywhere, customers, etc.
Cloud computing - rack space at Amazon.

Design team

CFO

Research

HR

Sales

Anything you do you can outsource all this stuff
$154 B.

Tiny tiny companies -

Outsourcing = total global market of everything.
panjiva

Outsources everything. Takes data in. Knocks data together.

Find someone who makes pants.

Find reputation

Who they sell to

Are they specialized

That is for COMPANIES
For individuals,

personal assistants.
Freelance marketplace

http://www.elance.com

http://www.guru.com

http://www.odesk.com

http://www.getafreelancer.com

http://www.rentacoder.com

http://www.j.gd/15901.html = Crowdspring --> logo, brochures

http://www.quirky.com
The crowdsourcing Process 8 steps:

1. company has a problem

2. Compnay broadcasts problem to crowd and asked for solutions

3. Online 'crowd' is asked to give solutions.

4. Crowd submits solutions

5. crowd vets solutions

6. company rewards winners

7. Company owns winning solutions

8. Company Profits
Wikinomics (3 years old)

Principles

- being open

- peering

- sharing

- acting globally

Put all the information about products on the net.

How many deals in last year.

How much time to deal.

So totally transparency and some of them are embarrassing, could take 3 weeks to 1 month to look at a deal.
Willling to open up, get people who know what they are getting as customers.

You know exactly what you are getting
Being open to peer to peer, sharing

The last ten deals have all been shared with other angel groups
Act globally, whatever you're doing, must look at a global level
Concept of FREE
What cannot be shared?

Virtually nothing
Jason was foudner of www.PriceLine.com

http://www.google.com/finance?q=NASDAQ:PCLN
Take the idea of globalize, sharing on national security

We have a big problem with lots of facilities and installations around the US that could be a potential target

But it's cheap with advancing technology, put cheap webcams on areas that should be no-man-land

take a shot every few seconds and outsource to crowd who looks for people in pictures
Do you see a person or vehicle in this image?

Options: Yes/No/Not sure

If someone sees a person more pictures are taken quickly and those pictures are crowd sourced until enough people say 'yes' and a security force is mobilized to intercept.
http://www.csoonline.com/article/218490/us-homeguard-someone-to-watch-over-you

\"Jay Walker, founder of Priceline.com, has moved from the auction business to the security business with US HomeGuard, an idea that would pay ordinary citizens $10 an hour to stare at surveillance video looking for terrorist activity.\"
Bob Rice - Three moves ahead

on Amazon: http://amzn.to/9lIBnA

Use heuristics like a human chess player in business to get great results, have solid ideas (protect queen, control corners and diagnals) and plan strategy around those

\"social networks will become profit networks as they are increasingly used to form jazz bands to attack specific tasks\"
The essence of entrepreneurial talent in the future

The ultimate entrepreneur of the future is someone who can pull talent around his area
Dave Brubeck

with your molecular structure, your talent

you should be able to form your own jazz band
Q: What DON'T you outsource.

One thing all this requires is the entrepreneur.

The unreasonable man. Shumpeter.

http://en.wikipedia.org/wiki/Joseph_Schumpeter
Q: Would you invest in a company where everything was outsourced?

A: absolutely, Why inhouse when better outsourced?

Plenty of fish site in Canada. Was one person. Now 10. Multi-hundred dollar corporation.
[web ref] Free Online Dating at Plentyoffish.com\u2122

100% Free online dating and matchmaking service for singles. 1300000 Daily Active Members.

www.plentyoffish.com/
Q: Everyone could copy your ideas if you're outsourcing them?

Keep constantly innovating. Pulling things together.

The entrepreneur must be flexible and must be constantly changing.
Q: Lifecycle shorter for outsourced business?

Hosting on amazon: nothing says it will be better hosted internally.

Just-in-time delivery/production

Can be left with infrastructure.

We are in a rapidly changing world - better to outsource.
Q: Back to crowdsourcing. Social and ethical implications. Crowd is unpaid, takes risks.

Issue raised with CrowdSpring. Less with

Are you exploiting people? Changing rules as Romer said.

Nothing wrong with working on spec.

Person is making the decision - time is worth it, on balance.

If Madison Avenue, $50,000 logo - you can't do it.

If your entire family's income is $50 / month - and you double your income. Not a negative.
Q: That's how architects work. One wins, rest waste several $1000.

I'm a market capitalist. Market will stabilize.

Scale whole thing. Look at economics.

Individual designers can sit on beach you saw in Paul's thing, that works.
Q: Pushing complexity - of simple tasks. What about complex tasks. Our company outsources, but there are always issues of quality and piecing things together.

True. We are in the middle of the curve. Now you are doing it. Challenges in quality and logistics. Reputation systems coming.

Pre-quality control

Outsourced project management

Mechanical Turk (again)

Anybody who needs something done - \"Is there a person in this photo\" to \"Review restaurant in Mountain View.\" Anyone can do it. Pay for it.

Social Mom. Backend platform. True blood story (again). Now on BluRay. Can connect to internet. Log in to Facebook account. It will tweet that you are watching and [yes, we know] it will make you look like a vampire in your photo. Sends your picture to MT to find two eyes and corner of mouth. They go \"beep beep beep\" and they get 25 cents.

Either bright 10 year old kid or someone somewhere far, far away in a distant land.

"

http://su-etherpad.com/core-Jun29-NCS-CL3

"Tuesday

6/29/2010

\"11:30 \u2013 12:30pm\"
Brad Templeton

NCS CL3 Network and Information
Slides: http://www.templetons.com/slides/singu-security-priv.ppt

All Our Ideas: http://www.allourideas.org/gsp10ncscl3
10:52

Salim: We are getting started. Please grab your seats.

Student with a brief announcement.

: VECOY meeting tonight.
Salim: OK. Red/Green.
Brad: Green - if you agree.

Can just shout GOOAAAL if you need to.

Our teaching fellow is from Spain.

What technology is going to give us.
Jan: [picking up at slide 37]

About what happens with software in this new regime.

All the other cores sit idle.

Database servers have already been written to use lots of cores.

Parallel query plan for SQL.

Web servers all send queries - peel off and send each request to a core.

Emergence of services in the cloud - divide up computation, they divide up, massively parallel, will scale beautiful.

3D graphics, games.

iPad can do molecular renderings.
Another place working well - technical and high performance computing.

Libraries off the shelf.

With GPUs and clusters, that much more headroom - 10s of TFLOPS in you office.
How do we more the mainstream desktop software so it can take advantage.

When I worked on problem at Micorsoft - mission statement:

\"Provide loveable parallel programming models, tools, and infrastructure to enable any developer to write robust software that scales up on new hardware.\"
Matrix example.

Sort example. Divide and conquer algorithm. Recursive split - on a large piece of data to sort, we are willing to go 10 or 100 way parallel if that many cores are available.
Stream example. Should scale as more cores are provided.

None of these are robust - require shared memory.

Each says, \"It is OK with me if you run this. Promise of no concurrent access.\"

Dominant model. Shared memory. Locks control shared memory. Fine for experts in small teams. Not good for large software ecology. No data race flags.

Some experts will still go to shared memory, more isolated models will be transitioned to.
Amdahl's Law. Remaining serial parts will still be slow.

If just 10% serial, won't get more than 10x faster, regardless of number of cores.
Personal computer processing is \"bursty\" (comes in bursts)

If you are typing the processor is busy, if you are not, it is asleep.
Idea: functions put in libraries

Models for recursive decomposition problems.

Message passing models - works well at internet scale and also in memory.

Can have isolated processes that don't share memory. This model does not have the shared memory, data race pitfalls.
MatLab example - lots of array processing.
Parallel Legacy Software?

Ought to be a \"magic\" compiler.

Word has millions of lines of code with complex interactions.

Lot of investment in automatic parallelization tools, but only small point successes.
Serial dependencies - calling library - takes a lock, puts in a block - maybe yes maybe no.

Key press, mouse movement ... a lot of computation is done on that core thread.

Because there are so many layers - requires detangling single thread assumptions.

Better to just start over.
If companies chase new rainbows, reengineer stuff that is 20 years old.
More likely: compose new ???
Brad said emerging technology has demand question. Games, technical computing. Not superhuge markets compared to PC and phone markets.

\"My mom\" does not need TFLOPS. Extra $200 does not help her - until those applications emerge. [Open Allure???]
New software gobbles up hardware and brings it to its knees.
In summary:

End of rapid scaling. Still 10-15% per year.

Software that scales at this rapid rate will have to be written in parallel.

Software has to be more efficient.

Need to have systems performance-optimized to become more energy efficient.

\"I reduced the energy by a factor of 2 - giving more battery life.\"

Will give us new opportunities in the decade ahead.

11:19
Q: Matt: Feasibility of 3D printing for circuits.
Jan: Couple of things. 3D integrated circuits are FOR SURE the future.

Another memory company doing stacking for density.
Also difficulty of scaling lithography. Contact printing. Not sure how they make blank, but they impress -- just like old lithography -- get parallel rendering.
Brad: ????

Jan: Maybe. Still like photolithography myself.

Maybe 3D printing - I should have gone to the talk last night.

We do 5 - 10 layers now. So 3D would just be more layers.

FPGA's - like a reconfigurable computer - different sets of gates that you can build at home. \"It is equivalent to the rule that
Q: No home ..

Jan: My hobby is building processors at home.

You can do anything in the garage if you have the right garage.

FPGA - I want an AND gate to be the input of this thing.

They've gotten huge.

A billion gates.

You could build anything Intel did 10 years ago you can do at home.

In addition to hobbyists,

specialist processing for chess, DNA,

you get something much better than general purpose chip.

Closest to hobbyist computer design.

Also people doing printing - inkjet printer transistors ..
Q: Emiliano: Hardware guys doing their things. CUDA. OpenCL.

Software guys trying to build new languages.

In the middle, what is basically missing,

PicStream?

Rapid7

Anyone else in the middle.

What is problem matching up?
Jan: Computer architecture folks taking top down approach. What hardware do we need.

Much better than past. I know a cool structure, maybe some application would light it up.

Bit late to the party.

Can't tell you about any favorite software tool startups.

Best work in parallel tools is Intel.

They have Parallel Studio - addresses seveeral considerations. Finding races. Uncovering parallel opportuniteis.

In addition to commercial projects, teams working

APL?

Microsofts, Intels have a lot riding on parallel scaling up.

apple OpenCL

Grand Central Dispatch
Brad: One more. Connor will host talk to help. Maneul will also help define terms. Jan here Wed. And me.

Connor: 9pm. No overlap with HIV talk.

Parallel computing, optimizing energy. Leverage large distributed computer networks created by cell phones. Push computations to distributed network where energy is cheap.
Jan: Like peer-to-peer computing idea. Not so clear about all the cell phones in California.

Not digging the energy market idea. Interesting, but sceptical.
Brad: Alright. Thank you very much Jan.

[applause]

11:30
Brad Templeton - Computer Insecurity and Privacy

Slides: http://www.templetons.com/slides/singu-security-priv.ppt
EFF http://www.eff.org/

Computer security sucks - yet we seem to be getting along

Most users are good, small num of bot nets and spammers

Automation automates both good and evil things
Firewall Hoax - like what we have at Nasa Ames

Isn't secure if laptops go into public places, get infected and then go back into their 'secure' homes or offices
Monoculture creates attractive targets, windows is so prevalent therefore most attacked

Liability for software makers won't work because one super virus will simple destroy the company

We haven't yet found how to incentivise greater security.
Bot Nets

Computer is secretly controlled by foreign user

30% of computers are botted

Primarily benign things like spamming

Most viciously used for DDOS (distributed denial of service) attack, could be used for extortion, costs about $8
Public Key Encryption - publish the key so everyone can encrpyt but only you can read, or everyone can read but only you can sign

- like a mails slot, creates a certificate/signature

- most comp security is based on keys

ZUI zero user interface - works so well no one knows it's going on. An example is Skype.

SSH - secure shell, used in command line, just works

SSL - it's a little involved to setup so 99% of web traffic is going on unencrypted wires
Example SSL Search - Google https://www.google.com

\"With Google search over SSL, you can have an end-to-end encrypted search solution between your computer and Google. This secured channel helps protect your search terms and your search results pages from being intercepted by a third party. This provides you with a more secure and private search experience.

To use search over SSL, visit https://encrypted.google.com each time you perform a search. Note that only Google web search is available over SSL, so other search products like Google Images and Google Maps are not currently available over SSL. When you're searching over SSL, these properties may not appear in the left panel.\"

http://www.google.com/support/websearch/bin/answer.py?answer=173733&hl=en
Phishing & Social Engineering:

http://en.wikipedia.org/wiki/Phishing

http://en.wikipedia.org/wiki/Social_engineering_(security)

A lot of hacking is social engineering, tricking people into giving you their passwords - (The Art of Deception by Kevin Mitnick)
A big reason 99% of traffic is not encrypted is because govt. made it a little hard to do and used weak encryptions
DES http://en.wikipedia.org/wiki/Data_Encryption_Standard

WEP http://en.wikipedia.org/wiki/Wired_Equivalent_Privacy
We'll learn more about quantum encryption which physics says is impossible to break
Some Topics:

Value of Privacy

Threats to Privacy

Now and in the future and other countries

Erasure of 4th amendment

Ease of use and user choice as negatives

Privacy is freedom.

When you are watched, you feel less free.

Heisenberg uncertainty

Observed that you couldn't observe a system without affecting it in the quantum physics world

A watched populace never boils

Apes (us) are the only ones who need privacy
Blinded By Science

Look hard enough in a big enpough sea of data; You will probably find whatever you're looking for; A seattle firefighter learned this the hard way; His house burnt down and his wife died; He was charged with arson and almost went to jail (because he bought a fire making chemical - for bbq - some time back), luckily the real arsonist was caught.

Scientists barely know this, juries definitely don't
You don't care about privacy until after it's invaded

You must protect others privacy to protect your own

There really is a slippery slope

We don't want to walk the edge of a police state
Shy people need privacy in a way that extroverts won't understand
Cloud applications

- Storing applications in the cloud - don't have to worry about distrtibuting it and making it cross platform complient and vairous individual settings on various systems

- Roaming, Scalability
[Slide of \"cloud\" - of atomic explosion!!!]
Pendulum effect

went from timesharing to pc back to timesharing (with clouds)
Consequences:

Data is leaving your hands

What we use to call spyware is now ubiquitous

You give up your right to refuse search and seizure when you give your data away like that, in effect you erase the 4th ammendment

We must be careful in not building the infrastructure of a police state

Don't intall the switch and make the question of freedom a policy question, by doing so you can have a police state without the tanks

e.g. Bush asking NSA which asked phone companies to tap the fiber optics due to terrorism without asking judges, later pardoned via POTUS working with congress
AI Privacy Invasion

- always been able to followable one person but now it's scalable to follow everyone and with AI we can have intelligent detectives following everyone
AI could judge us based on our past

- storage is cheap and laws change

- \"are you or have you ever been ...\"
Somewhere in the world governments will use facebook to roundup groups of people

- Don't facilitate big brothers

http://en.wikipedia.org/wiki/Expectation_of_privacy
China has pushed the Police State button.
http://en.wikipedia.org/wiki/NSA_warrantless_surveillance_controversy
Whole new level of privacy invasion coming with AI.

AI will have better ability to recognize faces, voices, will learn a lot of stuff.

Scalability is the key to everything.

Always been possible to follow people. It costs money.

Usually we mean it scales linearly.

Computers can do things a MILLION times more.

With AI, you can follow everyone with a \"Sam Spade\" (fictional detective)

http://en.wikipedia.org/wiki/Sam_Spade
ATM recorded your visit with a webcam.

AIs will scan the past, punish you for sins of the future.

Not only what is the threat today, but what can the future do - reaching back to hurt us.
http://en.wikipedia.org/wiki/Falun_Gong

What if Chinese Falun Gong followers had been on Facebook?

We do not want to be the preferred vendor for Big Brother (reference to Orwell 1984: http://en.wikipedia.org/wiki/Nineteen_Eighty-Four )
Traditionally we have had a balance.

We are changing that balance. Think about it.
Ease of Use is a Bug

Magnetic stripe on a license is an easy way to transer info

Before long web forms made you never want to join a website,

Facebook made it easy to share private info with fb apps
Must consider the impact on privacy and policies caused by these accelerating technologies
BEPSI: Bulk Export of Personal and Sensitive Information.

http://ideas.4brad.com/rename-data-portability-bepsi
Q: Alison: RFID -

http://en.wikipedia.org/wiki/Radio-frequency_identification

Brad: chip in your passport. No battery, powered by radio burst.

Can't check into hotel without showing your passport. May contain health data.

Can be read from 3 feet away.
Q: Sasha.Privacy Insurance.

Brad: About calcluating risk. Don't know enough.

Privacy insurance:http://www.networkworld.com/community/node/26132
Q: /France: Give prints at border. Took passport and tested it to see if it was fake.

Brad: At border, they can do unconstitutional things to you.
Q: National Security issues.

Brad: They don't inform us. Aware that security in the US is in terrible shape.

Idea that there might be a cyber attack.

Would like to have the internet privately owned. The way it is now.
Q: Steve Cronnin from the US. Looking forward 20 years, if government is capable of misuse. What protections can be put in?

Brad: Sceptical.

Proposal - number working on same philosophy. Own personal cloud. Like a safety deposit. Code comes to data rather than data coming to the code.

Google does not want to destroy value of data.

In Europe, there are laws on data retention. Some FORCE retention. Other laws FORBID retention.

doing everything with one company is monoculture.

I like Google a lot, but there are negatives.
Q: Even good intentions at corporate level doesn't stop one employee who could access data. Censorship in China.

Brad: They (Google) found China used their control to attack (dissidents).
Q: Rosa: Stockholders. Information is most valuable asset.

Google interesting in that way: give stockholders less power, keep it with founders.

You are trusting the founders of Google.
Q: Candice of Canada. Global privacy. International privacy push? Like human rights.

Unfortunately, anti-privacy push. [needs better link here]

http://en.wikipedia.org/wiki/Anti-piracy
Q: People who want access.

Brad: A little bit. Not up on digital rights stuff.
Q: David H/NZ: Next revolution DNA knowledge. People give up privacy out of complacency or information overload. What if there are values to it?

Brad: I asked for people who design these policies to make conscious choices.

I just want people to not go gung-ho and not consider the balance.
Q: Eric. Singularity coming. Know everything about everyone else. My DNA. My data. Just someone who wants to know it can do it.

Brad: Lot more surveillance that could be done.

Recommended the Book: David Brin, Transparent Society. Is on shelf from Smart.

http://en.wikipedia.org/wiki/The_Transparent_Society
Sarah: Anti-sedition act as knee-jerk reaction to war on terror. Cleared within a week. Public expression that questions government in any way was seditious.

Brad: Why didn't they fix it?

Sarah: New privacy laws for system which are almost as strict as Chinese government. We are blocked from speaking out against this privacy law. Orignially for children's rights. Accessnow.org.(http://www.accessnow.org/) Global organization to protect imprisoned bloggers. Will go in, if someone is arrested.
No independent voice. The threat of hive logic is so overwhelming.

Brad: AUS has great vision of freedom. Many did things equally bad. US had Patriot Act.

http://en.wikipedia.org/wiki/Patriot_Act

AUS has net filtering law

http://en.wikipedia.org/wiki/Internet_censorship_in_Australia
Q: Julielynn Wong: Canada/US. Is security of personal health information an illusion.

HIPPA is about security.

http://en.wikipedia.org/wiki/HIPPA

Have access to your health data but let the storage be owned by YOU not by a company storing it for you. We need to find a halfway between the PC we have and the CLOUD we are going to.
When databases are built, they have a job of saying, \"what can we do that is interesting\" and \"how can we make money.\"

If the company gives stuff free - the exchange is they get data to make money.

There is a pony in there (US joke reference: http://mustbeaponyinthere.com/ )
Q: Eugenie/France:

Brad: May deliberately set up system with no access.

Every attempt to anonymize data, can be de-anonymized because there are things in there to figure out who is who. Google's removing last octet is not enough.
Afternoon off.

Salim: As you can see, we have a fascinating set of discussions ahead of us.

Brad: Google thing tomorrow.

Salim: Justina is a postdoc at harvard.

Justina: personalized systems. From GSP09.

Em: camera lost
Day/afternoon off. Bus wil leave at 583C. to Shoreline park. Trails. Biking trails - we can borrow bikes. Frisbees/ toys. Til 5:30.

For the evening - another session sponsored by FEE track. Paul Roemer.

Other announcements.

For cards, please keep them in your badge holder.

Feedback for lecturers - already up in GSP10 portal. You can do that.

Another petition for opening up the gym.

Petition sheet in 583C.

Jose? Marathon.

1/ Organizing energy trip for one of two nuclear fusion. Other in france. We need 45 days of clearance. I need your signature.

2/ Etherpad for marathon: http://su-etherpad.com/sfmarathon
Q: Chiara. Gym = swimming pool.

Same people.

Brad: John Gage and I still working on track. I live in Silicon Valley .. if you say, \"my god, if we could go there ...\"

Salim: Two last things- camera - you guys are absent minded people! Great to have revolving policy. Going crazy with submissions.
[showing of Etherpad video]
[End of presentation]

All Our Ideas: http://www.allourideas.org/gsp10ncscl3


"

http://su-etherpad.com/core-Jun30-FSF

"Thursday

7/1/2010

11:30-12:30

Paul Saffo

FSF CL Cultivating intuition: effective forecasting in the face of rapid change
Slides:
Kathryn will review later looking for any [ref] markers--She and TFs will fill these in with references, but all are welcome (encouraged) to add references / put in as many details as available.

11:32

Paul Saffo - responsible for Futures Track here.

http://www.saffo.com

Smaller than last year.

Difference between catholics and congregationalists.

Catholics go to heaven. Congregationalists go to a meeting about going to heaven.

Entire summer for you is exercise in forecasting.

You have domain experts who are forecast minded.

In that spirit -
I teach forecasting to engineers. Have been here for 25 years in Silicon Valley.

I have outlived my own forecasts.

Just started a new company. Caught enterpreneurial disease from Peter.
Having the luxury of all these experts. Talking about specifics. What are the (heuristics) methods I use.
Forecasting Methods
Navigating uncertainty \u2013 effective forecasting in an uncertain time
[Slide CNN Hunt for Bin Laden. Experts agree leader is dead or alive.]

Very good forecast.

Reminds you of that lament: one armed economist,

I would wish for a one armed econoomist because they say, \"one the one hand ... on the other hand\"

Essence of forecasting is to embrace uncertainty

When facing uncertain situation, don't pretend there is no uncertainty just because it makes you feel uncomfortable.

Learn to live with uncertainty

Thus this is how I think of forcasting

You are thinking of an event and you can draw out the outcomes, not unlike a forcaster thinking about a hurricane's path, the same is done with technology

the cone going out from the event is narrow but it gets wider with time
Forcasting is not only difficult it is logically impossible, the fact is that the uncertainty behind forcasting is intrensic

I think that's good news

of course if we could get a peak into the futre, why bother
The bottom line behind why ones forecast is to get a range of possible outcomes and what can one do in the present to make the most agreeable choice happen
As you thinks of future lectures and team projects, always assume you're wrong. Forcast often and repeatedly. The interest is finding a cone that has balance, not too narrow not too wide
Try not to miss possiblities. Identify the key driving factors.

What are the driving forces underlying the thing that you are looking at that and are going to have the biggest factors in driving its change.

Jim Dater U of HI [ref]

Three kinds

*Continuations

*Cycles

*Novelties.
Constants:

Moore's Law. Predictable rate, we take for granted.

Imagine if \"will stall\" we'll be stuck with Windows 7 forever.
Usually the bulk of what is happening. We tend to miss because we take them for granted.

Continuations - identify them and make sure you identify all of them
Cycles:

Not here today, was yesterday, will be tomorrow.

Sun spots - starting cycle 24.

Kondratiev wave [ref]

In 1925 . Super cycles.

Only detail - it can be dangerous being a forecaster. Made Stalin very unhappy.

K died in a gulag in 1948.

Sometimes just tell boss what he wants to konw.
prosperity, recession, depression, improvement wave

We REALLY overlook the novelties. No frame of reference with which to measure it.

G basketball video. Don't tell - you miss the obvious.

Our perception causes us to miss things that are completely new

The mother of all novelties in 2008/9 not sure exactly when it happened human population is majority Urban.

In the 10,000 years since we started building this planet it never happened

Not like it never happened Lately - it neer happened AT ALL. Could have huge impact.

Was just a little blurb in teh news.

Dig underneath it = over 1 bn urban dwellers live in jury-rigged houses on land they don't own.

Hmmm maybe that's a novelty that could have a very big impact.

Once you've identified driving forces, look for Cross-Impact.

Rarely will one driving force be alone, look at more than one things coming together

Interaction among driving forces.

Cross impacts.

Take forecast of global population - humans can't grow exponentially foreever.

Inflection point due to urbanization.

Can we build livable cities or will we have a global slum.
Or does it become the global slumHam operators? Sun spot cycle example HF care, rest should as well. If really bad, our computers might not work.

Devices sensitive to electromagnetic pulses.

Sun spots may be a lot more relevant.
Scenarios - plural. If just one, it is not a scenario.

Explore options

look at different options

Back of envelope coctail party to million dollar study.
Multiple possilbe outcomes.
They way to integrate these all together is scenarios. The whole point is to explore and look at many different outcomes.

Here is how I approach it at its essence.

A scenario forces you to look at all the possibilites, cross impacts
Singularity

It is coming.

Hmmm. Are we sure?

What are the driving forces?

Tech change.

Other axis. What could I choose.

Public accpetance is assumed. - Maybe tehy won't accept it.

Four squares you can fill in.

Graph of Innovation Rate (stagnation to acceleration) vs. Public Acceptance (low to high)

I find 2 dim is enough. No compute rsoftware required.

High inv rate, public loves it. K's duet.

Shorthand for celebrating 150th birthday, playing piano duet with his late father who is present as an AI.

Come up with snappy titles - otherwise it blends into gray goo.
Singularity? What Singularity?

Just like space program. Went to moon. Expected Mars. Now \"pissing around in low earth orbit\" you can lean out the window of ISS and trail your fingers through the atmosphere.

Only geeks care about space today.

Rest of world could care less.
Not WORST possible - one step further.

Fast growth in tech and people gets scared. Mad as hell.

Remembe what happened to Knod

Public comes after you at night with flaming torches.
Final square:

Tech stagnates, but public buys in.

Desperately Seeking Singularities - people get mad. We counted on you to save the world an dyou failed us. Won't buy stock in your company.
So now let me ask you a quick question:

Show of hands. Who thinks which is most likely.

Other state?
Don't take your scenarios too seriously.

General public will not see what is coming.

you will confuse desired future with actual future.

You have to engage in Active Cognitive Dissonance to see alternatives.

The other reason to use scenarios is there is a

Big difference between forecast and reality.

Reality does not have to be \"believable\"
You see guys with a kite. They fly in it. suddenly people predicted personal flight.

Orville says, \"No. Never will happen. Only insane hobbyists wil have that. Instead, we will all fly around with aluminum cylinders with 100s of people and they will be unhappy when they are 30 minutes late.\" He would have been locked up.

Wing span is longer of entire flight of Wright brothers in 1903.

Think of Wildcards. Extend and test your scenarios.

Probe outer end of that cone of uncertainty.

Nothing is ever certain. 2012 Maya prophecy [ref]

Even, not knowing that, omit.
Unlikely things - reality has way of making unlikely come to pass.

Other contexts - things can become drivers.

You may feel an earthquake. USGS has done studies.

I have a high precision point on my property. It is moving 2 1/3 inches a year to the NW.

62% chance of quake.

Wildcard - magnitude 7 quake behind now and lunch.

If it hit, we've got the big one.

Driving forces can be wildcards if put in a specific time frame.
You know all about ...

from Ray's book.

Flat spot on left - pay close attention

We are linear thinkers. Very hard to forecast.

This tends to be what happens with visionaries.

They have such clear sight, they draw a straight line.

Early lesson I learned.

Never mistake a clear view for a short distance.

Entrepreneurs constantly do this.

Overestimate short term, underestimate long term.

People walk away just as business is taking off.

This can be very powerful tool for inflection point spotting.
What's going to be the next big hit? Look for something that has been failing for 20 years.

Rule of thumb 1: CHERISH FAILURE -especially if it is someone else's failure.

S curves apply to more than just tech

It can also apply to things like exploration of the new worl by europeans

Where did Columbus fall on this curve he was not at the turning point, he wasn't the first there, the first to make it back

Columbus came back when geo political environment was just right

Timing is everything

In biology innvoation is mutation and generally only happens when organism is under stress.

It's a very risky maneuver, Innovation (mutation) is generally lethal.
Habitat was an early Second Life

Rule of thumb 2: Indicators precede inflections (and the stranger the better)

1. contradictions - gold and google stock example Oct. 2007 vs Nov 2008

2. inversions - drug cartel cops example

3. oddities - roomba and DARPA grand challenge example
When times are bad gold is high stocks are low, when good stocks are high gold is low but mid 2007 - early 2008 both were high and it scared me

Rule of thumb 3: When change clusters at the extremes ... much more change is coming

Sure enough the economy tanked
Inversion of cops in hiding while criminal is proud in front of the camera, inversions happen all the time
When super wealthy is saying the roomba changed her life, people are naming bots and taking them on vacation, that was an inflection of robots making a big impact
Darpa Grand Challenge 1 - non completed, furthest got 7 miles, race was 150 miles

Darpa Grand Challenge 2 (one year later) - 5 completed the race
Reaching inflection point

Pile up of hundreds of cars north of Darpa challenge on the same day - people shouldn't drive!!
A new technology arrives and sit sets the space for innovation, for the take off of an ensuing innovation.

micro processor leads to PC

lasers lead to communication access/world wide web

sensor technology leads to interaction and robots \x3c- the next big thing
If you are going to look forward look back as well
Technological Ages

1. Chemistry - ealry 1900s

2. Physics - 1930s

3. Electronics - 1960s

4. Biology - 2000s
every technology works off each other incrementally

as you are moving forward think about these longer slower fields
What can you get a date with at a cocktail party?
Tip Jar: If you fear change, leave it in here
Embrace Uncertainty! But keep in mind you will scare everyone around you.

paul@saffo.com
[end of presentation]
Watch 2 minute Tech Update on YouTube:

http://www.youtube.com/watch?v=EyNBZMjz2QU

on SU video server:

http://video.google.com/a/singularityu.org/#/Play/contentId=12569090359a5e04

=================================================================

Question Template [for copy/paste - do not use]

Vote by adding +XX where XX are your initials (see key: http://su-etherpad.com/initials)

Student Question:

Student Votes:

Student Comments:

Faculty Response:

=================================================================

Student Question: http://www.saffo.com/essays/disinteremediation.php says value chains are getting longer, not shorter. This came up yesterday for me as I looked at bicycle sharing which has become successful due to an ADVERTISING company. What other examples / opportunities for longer value chains do you see?

Student Votes: +JG

Student Comments:

Faculty Response:
Student Question: What comes after bio?

Student Votes:

Student Comments:

Faculty Response: Perhaps nano, even though that's combined with bio

History of technology. More control over smaller stuff.

Nano.

Student Question: Black Swan [ref] How have you dealt with that in forecasting?

Student Votes:

Student Comments:

Faculty Response:

An indicator, read book. Nice meditation on how unlikely events can be much more likely.

Lebanon was place he learned to expect unexpected.

Systems susceptable to
Let me recommend:

AI Track. You've got to read this. Software robots here already. Scarry consequences of bots. Daemon by Daniel Suarez.

http://www.thedaemon.com/

Student Question: What things are people excited about that probably won't pan out anytime soon?

Student Votes:

Student Comments:

Faculty Response:

Has anybody taken a Roomba apart? There probably won't be a all-in-one AI robot anytime soon

It is just a pile of transistors. It is an imposter.

Vision of intelligences smarter than us ...

Most of our friends can't pass the turing test.

Good news . it is probably far off. If we are very lucky, they will treat us like pets. If we are unlucky, they will treat us as food.

Student Question: Backcasting. Merge with forecasting
Eric Eze: Backcasting: http://en.wikipedia.org/wiki/Backcasting

Eric Eze: backcasting 2: http://www.thenaturalstep.org/the-system-conditions

Student Votes:

Student Comments:

Faculty Response:

Start wtih preferred future. Work back. Very powerful tool.

Thanks for mentioning.

[applause]

Salim: Couple of announcements.

Different format to try more interaction/discussion. We'll try that out.

Bus to Google -your name has to be on the list for you to go.
Yoga and basketball and boot camp tonight.
Canoe and beer for Canada night.
Instant Evaluation:

https://sites.google.com/a/singularityu.org/gsp10/feedback/week-2
"

http://su-etherpad.com/core-Jun29-NCS-CL2

"Tuesday

6/29/2010

10:15 \u2013 11:15
Jan Gray

NCS CL2 Network and Information
Moores Law: The Marvelous MOSFET, see slides with image.
Notes were just put at the end of the http://su-etherpad.com/core-Jun29-NCS-CL1 pad.
Jan:

Slides at https://docs.google.com/a/singularityu.org/leaf?id=0B6vjqTaW5opaYWU0MWJlODctZWQzZi00MTk1LThiN2UtZDlkODNjMWMwODAw&sort=name&layout=list&pid=0B2Ij9HPeinaFODJjNjVlYTItNTlhMC00YzI0LWFlZDEtM2U3NTg3ODc5OTNj&cindex=1

I was born when Moore's Law was framed. Hope it will keep scaling up.

Around 2004, ebb in performance scaling. Run from 1970 to 2000. Doubling every 2 years.
1 MHz to 3000.

Each clock cycle did more.

Heat disappated went up.

Rounding off of clock speed: 3 GHz

Yet we've gone from 10,000s to millions of transistors on a die.
do we have a good c?

architectures - how to spend them?

How does software change?
Some electronics. MOSFET

Transistor is three terminal device. Source, drain. Control gate.

Channel has semi-conductor material.
If NO voltage, channel doesn't conduct.

If positive voltage, you get conduction from source to drain.
[Video - sand to silicon chip]

Make one large silicon crystal.

Slice it.

Photolithography.

A resist. Expose through mask with many many copies on round wafer.

Etch, deposit.

Build parts of transistor - 9 layers of metal.

100s of saws cut it up.
Several hundred billion transistors on that one die.
[Chart from S. Borkar]

Each 2 years, we go into a new technology Node. Smaller by 0.7x. Doubling in 2 year.

Quadrupling of transistors on a die.
Quarter of a trillion transistors on a die by 2018.

Rate of speed improvement is going down. Not as well as they used to.

Then it will slow down to every three years after that
Extend 193nm refractive lithography

Optical proximity correction, double patterning
13nm EUV lithography

soft x-rays, vacuum chambers
Transistors:

new material in gate, insulator, channel
DRAM - capacitor, 2x/3y

FLASH PROM - floating gate 2x/1.5y 64gb at 25nm
3d: stack cells vertically

resistance change/phase change RAM

amorphous/cystalline
Slowing of transistor doublings (now we are 2x in 3yr)

But.. cost halvings still continuing - \"22nm ought to be enough for anybody\"
Computer Architecture

1986-2002, computers 2x in speed every year but now its leveling off
Instruction Level Parallelism (ILP) [cut and paste from slides]

Spending Transistors on Performance - run programs 2x as fast
L1 cache reduces memory access time

Branch prediction for determining where the program is heading

Vector operations FPU's for working with videos/graphics
Over 30 years - CPU cycle time 1000ns -> 0.3ns but DRAM access time 500ns -> 100ns

sometimes progress stops for hundreds of nanoseconds
\"The Power Wall\"

Over last 30 years: die size larger, voltage 15->1v, frequency 1->300MHz, Power 1->100W

like the heat of a rocket at 1000x and surface of the sun at 10000x, just can't go there
P porportional to CV^2f
the power wall is the limiting design constraint
\"Complexity Wall\"

diminishing returns in wide-issue machines

only a tiny part of the die is actually doing processes
\"Explicit Parallelism - Multi-Core\"

much more area efficient

power wall is still there, how to solve?

finesse power with lower V, lower frequency, sleeping of transistors

finesse memory wall with memory parallelism

can't think of anything better so let's see if we can get the programers to get it to work

32 cores = 1 trillion floating points per second
Single-Chip Cloud - 48 individual cores
\"Scaling Up Memory Bandwidth\"

TFLOPS need TB/s of data.... 100s of GB/s not nearly enough

solution: bumping, die stacking

\"Silicon Photonics\"

should surpass copper traces for non-local interconnect

optical links at TB speed

Die stacking + Photonics = adequate bandwidth for next generation
\"Graphics Processor Computing\"

100s of cores

1000s of threads

Graphics module is very accessible to scientific and technical computing.

GPU use is still \"black magic\"

Game enthusiasts buy $500 boards and drive supercomputing.

Nvidia chip with 500 cores.

http://www.nvidia.com/content/global/global.php

http://www.nvidia.com/object/cool_stuff.html
Systems on Chip, Network on Chip, Modularity, Heterogeneity

database with large cores, small cores, media cores etc... mix and match for different problem domains

After 2015, a lot more fault detection on chips

Eventually to the point that computers are modular/stackable

iPhone G4, optimized for low energy
Eurocloud - research project that makes 3D cloud computing on a chip
Break at 10:26

Brad: I hope this was good. [Green and red card votes.]

Building a chip today is like trying to paint art with a paint roller
Connor: SU exponential technologies. Age of abundance.

How will computer design / chip /datacenter design when energy is limitless.
Jan: All large scale production systems will be expensive, hard to deploy. Don't know where ..

IF we had unlimited energy, constraints would go away until planet melts down.
Brad: Unless we come up with technologies, like reversible computing.

They get as hot as surface of the sun.
Jan: We'll just use lasers to put energy in space.

Brad: Or put computing in space. [Dmitry?]
Eric: Glass revolution. Explain a bit more. People-to-people.

Brad: In English, Grass roots is something coming from bottom.

Markets have exponential growth. If networking is done at grass roots, connecting neighbors with Gigabit.

Like Wi-Fi - always a hotspot somewhere near you.

Sometimes give away free.

Revolution of access.

Take that model so you can go to local electronics store.

Fibre is safe.

Not worth stealing - unlike valuable copper.

Sends immense amount of data that does not involve phone companies.

[applause]
: Wondering about cloud. What is next?

Brad: Stratosphere. Proposal to put computers in space. Actually latency to LEO is quite good. As good as using something on your desk.

Other proposals - quantum computing next week.

Strange rules of quantum physics to do computing in small groups of atoms.

May still have clouds of quantum computers.

That will stick around.

Jan: 50x or 100x cost reduction of processors in the cloud.
Brad: Buy computer today, run for 3 years. What is most expensive part:

The ELECTRICITY.

Not fixing that well right now.

Home energy track may address this.
Dmitry: Vacuum of space? Cold? Problems.

Brad: Only cool by radiation. Not so good.

Dan Berry- when they take laptops into space, they must put fans on them because cooling does not happen by convection in AIR.
: So the cloud is built and operated by companies at the moment. Should it be?

Brad: Next session.
Brad (GSP Brad):

Jan: PhysX - software physics interface. Discontinued ?agea?

Now in GPU. Gathered they felt scaling of GPU did Physics stuff without using 2 chips.
Sarah/ Australia: Potentials for supercomputing involving biological systems? Interfaces?

Jan: Lovely image of chip + neuron.

Inspired architectures - neocortex type in hardware, very interesting.
Brad: DNA computing?

Very slow, but for free, DNA will grow and may come out with calculation.

Jan: Biological curcuits. DNA has nice property of complementary base pairs. Some parts of the strand bind tightly. Others left dangling. You can do AND gates. Strand activates whole other computation.

Not aiming for TFLOPS. Rather simple computation to unlock drug at particular site.

Kind of taking experience with computers and getting state, amplification in DNA strands.

Sky is the limit on what you could do with this.
Brad: You guys have been very interactive.

Tony / Korea: End of Moore's Law? Society is depending on it - some industries would close? Implications of plateau?

Jan: Every technology has one of these cycles. Fortunate technologies go through phases, slow down, turn to new technology.

Theme of this talk - uni-processor slowed, however with more cores - with parallel software we should get back to fast rate of scaling.

Not stopped growing - still getting 10-15% better. That is amazing.

We will either transition to something else or cheaper ...

Brad: Pause now. Come back in 7-8 minutes.

10:43
NEW PAD AFTER BREAK:

http://su-etherpad.com/core-Jun29-NCS-CL3
"

http://su-etherpad.com/core-Jul13-SPS

"7/13/2010

Tuesday

9:00 - 10:00 am

SPS CL1 . Exploration of Mars (Pete Worden)

Slides: https://docs.google.com/a/singularityu.org/fileview?id=0B2Ij9HPeinaFOGEwYmExODktZTAyOC00Y2E3LWFiM2QtYzZmODI4ZWZlYzgz&hl=en
10:15 \u2013 11:15 am

SPS CL3 Outward from Earth (Chris McKay)

Slides: https://docs.google.com/a/singularityu.org/present/view?id=0AWIj9HPeinaFZGY2cDNjOXpfNDQ3ZGI4anYyZ2M&hl=en
11:30 \u2013 12:30 pm

SPS CL2 Rocket Equation (Dan Barry)

(outdoors

9:10

NASA does three things.

1/ Science. Has been exciting decade. We don't know hardly anything about the universe. That's called job security.

2/ We do a lot of stuff to help life on Earth. Climate change. Revolutionize green aviation. Carbon neutral, eventually carbon free air travel.

3/ Human space flight. Objective is to begin settling the solar system.
What we are doing. Where we might go.
- Virgle. The Adventure of Many Lifetimes

http://www.google.com/virgle/

Joint venture between Virgin and Google called Virgle

On April 1st, but may not be a joke.
- Settling Mars

Most likely place for permament human habitat is Mars

Typically interesting part of mars is below the surface

Keys to settling mars or anywhere else

1. In-Situ resource utilization

2. stepping stones via asteroids

3. one way? \"president doesn't agree but I do\"
- Mars Surface - Showing crater with ice

Mars: It's Big, pretty dead world

Surface is volatile and rich
- Potential aquifers under the soil

- Methane on mars suggests potential life processes
Life on Earth could have come from panspermia, asteroids from Mars containing life

http://en.wikipedia.org/wiki/Panspermia

Near-Earth asteroids may be a good stepping stone for getting to Mars
-Known (current) NEO Population

There are thousands of these things that are easily accessible

http://www.cfa.harvard.edu/iau/lists/InnerPlot2.html
- Asteroid Itokawa

1/2 kilometer, typical asteroid between earth and mars
- Asteroid and Comets visited to date

Itokawa is just a dot compared to other asteroids

Composition ranges from almost pure metal to ices and other volatiles( bodes for life)
- NASA's exploration lauch architecture

Saturn 5 used for apollo missions, replaced with shuttle that turned out to be not so cheap

Underfunded by 50% - 200%

- NEO missions

-

Easier to go to an asteroid than it is to go to the lunar surface

Moon has about 1/6 gravity but asteroids are much smaller and have essentially no gravity

- 5 month mission to a near earth asteroid

- Exporation Metro Map

If were going o go to a lot of places in the solar system why not build something that can do that

A space metro system

Build sstems that can go to asteroids, mars, outer planets and equilibrium points between objects
http://en.wikipedia.org/wiki/Delta-v
- We want to turn low earth orbit travel to private sector

Later in the decade want missions to get to the moon

Build the first true space ship

Next step would be getting to equilbriums and space walks near these objects

Final step is to Mars

- Reusability and Advanced systems

using 'ion' drives
-Stepping stones strategy

moon and lunar robotics, deep space, asteroids, moons of mars, mars, ...

-

get to main belt in next [?] years

- Asteroids

building blocks of solar system

overtime collided and coalesced into planets
http://en.wikipedia.org/wiki/Tunguska_event
-

Nasa Ames is getting half a billion with DOD to create space ship [SPE?]

- Mars has two moons Phobos and Deimos

why to phobos? may contain volatile surface, raditation safe haven, can use robotics, good staging ground

-

-

Itokawa is prbbly, others are smooth, shows differences we need to do spectrography

- LCROSS \"bombing\" the moon

I twittered \"Nasa Ames is go for first bombing run of the moon\" got some conspiracy theorists worried and leadership told me to shut up
\"NASA Bombing of the Moon may create conflict with ET's, UFO's\"

http://macedoniaonline.eu/content/view/7216/56/
Cost 78 million dollars, pretty good mission

-

- Benefits of Commercialization

Falcon 9 launch in June, first rocket feveloped in US in a long time completely with private funding

-

Photo of illuminated via natural illuminescence cyano bacteria, one of the oldest organisms we know, helpd to create an oxygen atmosphere

-

- Biology beat checmicals and physical methods for food production

Cyanobacteria spirullina

- Craig Venter

\"Over the next 20 years, synthetic genomics is going to become the standard for building anything\"

- Applications of synthetic biology to NASA

- Pharmasat -1

http://tia.arc.nasa.gov/pharmasat/

On Twitter: http://twitter.com/PharmaSat/lists/memberships

- Harabusa NEO and Tagish Lake Meteorite

http://en.wikipedia.org/wiki/Tagish_lake_meteorite

- meteor is about 30% volatiles, looks like coal, can we grow cyanobacteria on these

http://en.wikipedia.org/wiki/Cyanobacteria

-

-Deimos and Phobos

Orbits are very close to mars, makes a fairly good space station for telerobotic exploration of mars
- Life on Mars?

say what we thought was fossilzed bacteria on asteroids but much smaller than any bacteria on Earth

- War of thw Worlds

Life is a scary thing, let's make sure there's none before we do any drilling

- Places in space for life

Meteors and Asteroids, Craters with Ice, Mars

- Terraforming Mars?

seems life the possibility is there

very interesting prospect

- Other target opportunities

-

Paper says there could be life on mars or Titan due to methane

-

Other worlds with liquids, Mars, Europa, Enceladus, Titan

-

Keplar, first device primarily used for finding earth like objects

we now have 6 months worth of data with about 20 candidate planets but there are other factors

-

Possilbe to find Pandora like moons

http://en.wikipedia.org/wiki/Polyphemus
-

Close with a few words on private funding settlement

http://en.wikipedia.org/wiki/Mayflower (didn't go so well)

- Pick of pilgrims

it's a serious issue

-

There are caves on Mars that could be potentially sealed and could be used to create earthlike conditions and atmosphere

-

Nasa toyed with idea of selling space but didn't go far

-

Now more progress is being made in commercial sector

-

In conclusion

1. We will soon have ability to visit Mars and other places in Solar System

2. Settlement is the only rationale for human space travel

3. Path to Mars could be via near earth asteroids and Phobos

4. Use bio-engineeed tech to make what we need

5. Private one-way setllement missions could occur soon
Q: With everything moving in space how much do launch windows affect these missions?

A: Does make a difference but not that much ???
Q. How well are space technologies being deployed on Earth?

A: Use to be good at this but not anymore, Nasa just doesn't have much better tech than private sector now
Technology Development and Infusion from NASA's Innovative Partnerships Program

Douglas A. Comstock

National Aeronautics and Space Administration

http://www.nasa.gov/pdf/324711main_ieeeac_paper_138.pdf
Q: What are the reasons that private rockets are cheaper now?

A: NASA had a lot of money and maybe Elon Musk

http://en.wikipedia.org/wiki/Elon_Musk

did use/steal previous work but so what? Highlights the efficiency of privately funded lean and narrow focused endeavors.
http://nasajobs.nasa.gov/
Invited to NASA bar and golf course.

10:04 [applause]

10:16 Chris McKay

-

Astrobiolgy is the study of the origin, evolution, distribution, and future of life in the universe

The future involves human choice and actions

-

I want to propose an answer to the question: why do we care?

-

-

Whiy is Mars useful for human exploration

1. Did Mars have a past in terms of life?

no role for humans here

2. Can Mars support life?

involves humans

3. Can Mars have a biologiical future

our decision

-

Don't send humans because they will contaminate Mars before we can determine it's own potential for life

-

True story, on July 4th 1997 we sent bacteria to Mars, they came unarmed and peacefully

Most died

They were dormant on the landing craft

Pathfinder

-

After Viking we stopped sterilzing landers

We know where the contamination is, we know where the parachute, heat shield and lander are

-

-

Death on Mars from UV light means that bacteria can't live on surface, we could remove contamination by cleaning up pieces

-

Rabbit removal from antartica vs. australia is linear vs. exponential for same reasons of harsh environment

-

Three possibilities for Past life on mars

1. No life, no worries

2. Common origin of Earth life, very little worries

3. It was a second genesis unrelated to Earth, true aliens, could be worries

-

Tree of life, all life is just one biochemistry, aliens could potentially not be on our tree of life

Assigning Moral Status:

- Moral Agents are rational,

- Moral Status, we humans assign moral status

Pain, complex behavior & communication, or membership in a set.

Set theory example: being human (if your not rational, you have same moral status as rational humans, just for being a member of a set)

2 sets we assign moral status: life and sets
-

Moral Status of Alien Microbes,

Argue that microbes score moral status on pain and ect.. but high moral status for life 2.0.

-

Argue we need to biological reverse exploration.

Drilling below the surface could lead to non-reversible bio contamination

-

-

We should have bio-reversible \"ctrl Z\" option

ok/cancel prompt: \"Are you sure you want to drill into thes subsurface Martian aquifer?\"

-

-

Do you notice the difference between these two pictures?

- Long term view of beinging Mars back to life, terraforming, Water may have been a marble, now is dry, maybe will be again in the future

Don't have all the components for life but Building blocks are there we just have to assemble them

-

Terraforming mars is live a descent down a high mountain

-

How do we warm it up? Use greenhouse gases other than chlorine and bromine
There is no meaningful difference between life and nature on earth. But in Mars there is no life, so do we value nature or life?
1990 McKay, tried to evaluate ethics independant of life on earth. (see slide)

My adding life, we can add it some sort of intrensic value.

Takes actions of Deep Ecology (Naess, 1984) as true, then
Deep Ecology:

There is inherit value in life, the more wide spread and diverse the life the more valuable.

Our goal on Mars is to enhance and enrich life in the universe.Deep Ecology:

There is inherit value in life, the more wide spread and diverse the life the more valuable.

Our goal on Mars is to enhance and enrich life in the universe.
Q: What about extremophiles

http://en.wikipedia.org/wiki/Extremophile

A: unique in location but not in DNA
11:03 [applause]

Dan: Next lecture by building 20.

Em: Announcement - badges.


=============== POST-presentation notes
Instant Evaluation:

https://sites.google.com/a/singularityu.org/gsp10/feedback/week-4
https://spreadsheets.google.com/a/singularityu.org/viewform?formkey=dGZadGpDejljY0JaWnRnV2Q1SEdYRGc6MQ
"

http://su-etherpad.com/core-Jul12-AIR

"

7/12/2010

Monday

9am - 10am

AIR CL7 AI Agents and Assistants

Adam Cheyer

Where: Ballroom (B3)

Slides: http://adam.cheyer.com/news.html (actually, link is broken ...)

Slides in ZIP file: http://bit.ly/aTOZRt
10:15 Dan Barry

AIR CL8 Robots in the Real World (Dan Barry)

Cooperation trumps competition
11:30

AIR CL9 Reverse-Engineering the Brain

William Risk
=============== PRE-presentation notes
Erez has introduced a new SUMMARY Etherpad:

http://su-etherpad.com/Pad-One-Liner:

Please contribute.
Slide marker in Etherpad:

-

at the left margin, the hyphen above indicates a new slide.
If someone wants to put in the TITLE of the slide, like this

- Example Title

that would be even better.
[ref]

If you need a reference, use this marker.

If you missed it on Saturday, July 10, Gregg Maryniak gave a second full presentation on Energy starting from line 693 in this pad:

http://su-etherpad.com/core-Jul10-EES

He also gave a third full presentation to the Space team on Sunday, July 11:

http://su-etherpad.com/Jul11-GM
Aubrey de Grey spent some time in 583C talking about SENS on Sunday afternoon, 11 July:

http://su-etherpad.com/event-Jul11-AD
PRIOR AIR pads:

Jun28 AIR CL1 Overview AI Robotics Jacobstein

http://su-etherpad.com/core-Jun28-AIR-CL1-Overview-AI-Robotics-Jacobstein

Jul5 AIR

http://su-etherpad.com/core-Jul5-AIR

Jul12 AIR

http://su-etherpad.com/core-Jul12-AIR
Spreadsheet of all core Etherpads:

http://bit.ly/GSP10library
Link to full SU Etherpad index:

http://singularityu.org/gsp10pads/
=============== End of PRE-presentation notes
Adam Cheyer

http://www.adam.cheyer.com/about.html
Siri founder

http://siri.com/

\"Siri understands what you say, accomplishes tasks for you and adapts to your preferences over time.

Today, Siri can help you find and plan things to do. You can ask Siri to find a romantic place for dinner, tell you what\u2019s playing at a local jazz club or get tickets to a movie for Saturday night.

Siri is young and, like a child taking its first steps, may be awkward at times. Siri may occasionally misunderstand things you ask it to do even within its range of understanding.

Nonetheless, Siri will improve quickly by getting to know you better and understanding a broader set of tasks. In fact, right now, Siri\u2019s learning how to handle reminders, flights stats and reference questions. Our vision is that, over time, you\u2019ll trust Siri to manage many personal details in your life - from recommending a wine you might enjoy to managing your to do list.\"

YouTube demo: http://www.youtube.com/watch?v=MpjpVAB06O4
http://www.change.org/

\"the premier social network for positive social change\"
http://www.linkedin.com/companies/sandwalk-capital-hk-limited

Sandwalk Capital is an asset advisory firm specializing in the development and support of systematic short-term trading Investment Strategies in the most liquid markets. We focus on advanced pattern recognition systems using proprietary evolutionary algorithms and a robust testing and evaluation framework to uncover new sources of alpha.

Slides: http://adam.cheyer.com/news.html
AI: What's Possible, What's Not, How Do We Move Forward?

9:09am Neil: [bio]

Siri recently sold to Apple

See free iPhone app (AI agent)

http://siri.com/
Author of more than 50 peer reviewed papers and 9 patents.
Adam: Really impressed. Kudos.

Updated version of talk from last year.

Questions:

- The Future of AI

What's missing?

What works well? Where are the gaps?

Where are we today?

What are the best AI engines - limitations?

What is the best hope of success? low hanging fruit?

Where can we break through?

Predictions ...
- video from 1986 - Apple - about 10 minute video

Really sets the bar for an assistant.

How close are we.

We have the touch phones

We have the internet

Interacting with human

Awareness of temporal and social context

What is happening in a person's life, how does that fit

Phone has all contacts

We have continuous speech both in and out.

Here we see conversational assistants.


Artificial Intelligence\u2028\u2028What\u2019s Possible, What\u2019s Not, \u2028How Do We Move Forward?

Adam Cheyer\u2028\u2028Co-Founder, VP Engineering

Siri Inc

The Future of AI

AI: What are we trying to achieve?

What does it require?

Why is it hard?

What approaches are there?

What works well?

What doesn\u2019t?

Where is the state of the art today?

Commercial \u201cAI Engines\u201d

What are our best hopes of success?

What holds us back?

What does the future look like?

In 5 years.... In 15\u2026 In 25\u2026

AI: What are we trying to achieve?

Interaction with the Assistant

Touch screens and cinematic animation

Global network for info and collaboration

Awareness of temporal and social context

Continuous Speech in and out

Conversational Interface - assistant talks back

Delegation of tasks to the assistant

Assistant use of personal data
Is the Knowledge Navigator vision \u2028possible today?
Is the Knowledge Navigator vision \u2028possible today?

How Close are we Today?

Touch screens

Cinematic effects

Global network

Location and time awareness

Speech out, on demand

Continuous speech to text
Our iPhone has all these pieces.

Being able to communicate, delegate, etc.

You just can't talk to a search engine this way.
Peter Norvig was here.

John Battel wrote - a conversation [ref]

Why is it hard?

Each component technology is complex
Not enough to have dialog models, task models. You need to put it all together.

Each component is quite difficult.
\"Book 4 star restaurant in Boston\"

The word \"book\" is an object, a city ...

\"Star\" is a restaurant name ...

There is a restaurant called \"Restaurant\" ...

It explodes into a huge possibilities ...

If you try to put the raw semantic interpretations into grammar.

Informal, incomplete grammar of English is larger than 1,700 pages\u2028\u2028

R. Quirk et al., A Comprehensive Grammar of the English Language, Longman, 1985.
What is the meaning, the intent. Especially when the words come in context.
Why is it hard?

\u201cCommon sense\u201d knowledge is fundamental to all components

Don\u2019t yet have sufficient representations for logical reasoning

*Huge* amounts of knowledge required, where does it come from?

How to manage the scale of the two?
\"Journal articles only\" set of results too large.

All of that knowledge would need to be encoded in a computer.

What is the right representation for putting in these facts and rules.

Huge amounts. How do you encode in a reasonable way.
Each component area uses different technologies, languages, methods yet \u2028deep integration is fundamentally required
You can't do this at a superficial level. DARPA spent $250 m. Funded stove-pipe research. Ontologists. Linguists. Planning specialists. Machine learning. CALO taught about movies? Make changes across entire set of technologies - words, rules, objects - very difficult to bring all the people together quickly.
- What approaches are there?

Simple heuristic rules plus enormous computation (search)

\u201cDeep\u201d knowledge approach

Typically relies on hand-coded grammars, ontologies, and rules

Statistical approach relying on learning probabilities from large corpora
Examples: Search will take Page Rank as one metric. Try to optimise on certain criteria. Look through entire set of data and evaluate on this criteria.
Can have very simple rules, but lots of power.
2/

Someone is going to code up knowledge. Autonomous vehicle. Not a search problem. Techniques were domain specific.
3/

Stat approaches. English and French - you just train. Run algorithms. Gee, when I say \"book\", then \"livre\" in French might be good.
-What works well?
All the approaches work well \u2013 for some problems
Massive search with simple heuristics

Deep Blue beats world chess champion

Genetic Finance beats benchmarks on stock prediction

Massive distributed learning. Like SETI at home. Pretty picture shows and it starts crunching. Rather than look for intelligent life, we look at the stock market. Run on millions of computers. Which indicators are most useful is a massive search problem. They get exponential in size.
Statistical training based on massive data

Speech recognition

Purely a data problem. Here are the words, here are the phonemes. Let's match them.

Machine translation

Google has orders of magnitude more data. Google vs Bing Translate - data helps.
Web search

Read: \u201cThe Unreasonable Effectiveness of Data\u201d [ref]

Norvig - algo that work OK with 10 million make leaps with order of magnitute more data.
\u201cDeep\u201d knowledge approach

Urban Challenge/Robotics

Multiplayer Virtual Games

Someone has coded up rules.
- What doesn\u2019t?

But they have their limitations
Massive search with simple heuristics

Only certain problems fit into this category
Think again of the chess computer. Heuristic - 1 get the king, good position = better score.

In Genetic Finance, make money good, lose bad. Take transaction costs into account.

\"Bring up my rainforest slides\" will not fit this approach.
Statistical training based on massive data

Again, works only for certain problems due to availability of data and shallowness of scope
It is changing - we have video cameras on anything. Get data on how children learn language by hearing every single word they say. Question of getting data is a key lever on are we going to have really intelligent systems. Are we going to have systems to see and \"experience\" the world. In a sense.
\u201cDeep\u201d knowledge approach

Too brittle

Take text-to-speech. Chevaneski? [ref] Changed the world.

How would you pronouce these words?

Took a dictionary - found phonemes.

Took neural network and trained words input to phonemes output.

Started with RANDOM output, system learned to feedback, get better.

Played over time, sped up, you could here the concepts get better and better.

Want to learn Finnish? Just give me another dictionary.

Who comes up with rules is the challenge.

How to get the data?

- State of the Art: AI Engines

Google \u2013 A \u201cSearch Engine\u201d

Amazing.

Wolfram Alpha \u2013 A \u201cCompute Engine\u201d

http://www.wolframalpha.com/

Mathematica at the core

Huge, curated fact base
9:31 video

Brainchild of Stephen Wolfram [ref]

What he did was interesting and somewhat revolutionary.

Put a math engine at the core.

AI people said, \"What? It doesn't do logic?\"

[\"What is the weather in NYC? Current temp, and forecasts, conditions, historical temp data. Let's get even more specific. April 2, 1981. Great for research projects. Here is that pattern, daily breakdown of cloud cover. Type another city. Say Philly. On this day in 1981. So, it's thinking. Side by side comparisons. Very interesting. Do another search.

How much is 440 British pounds worth in US dollars .... over 10 years\"]

What Stephen has done is curated a massive set of data, formulas and rules.

Stephen Wolfram has curated massive ammounts of data, rules, formulas across all areas of science, wolfram alpha, a new type of engine that can answer new kinds of questions

Run in parallel across

Chem

Bio

Astrophysics

He says search engines only retrieve what is known. This can return answers to questions that have never been posed before.

He has 10,000 machines running Mathematica in parallel.

Just type \"6\" and you get all sorts of disconnected facts.

April 2, gives you the number of days ago, in parallel. These different math engines.

Very interesting system.

Who enters this data? He wants precision to the Nth decimal point.

The idea of Wikipedia is abhorrent to him.

He has developed tools to manage the quality of the data - the actual quality is one of the most important achievements. It will advance at the speed he can get his experts to curate that data.
True Knowledge \u2013 An \u201cInference Engine\u201d

http://www.trueknowledge.com/

Chains inferences together, reasons about time

Collaborative knowledge entry for enhancing KB
You can ask who was President when Barack Obama was a teenager?

First who is Barack Obama.

Answer - actually 4 Presidents during his teen years.

This is collaborative.

You could hire guys.

This is trying to create an application - if it doesn't know, it asks, \"Do you know answer?\"

[video \"Big problem - computers don't understand the content of websites. Just keywords ... \"Is Jennifer Lopez single?\" Represent knowledge in a format that computers can understand. Can be used just like a search engine. When you type in a question this time, you get a perfect in-line response. Disambiguated meaning of \"single\".]

skipping ahead

[\"Users can can contribute templates. Enter facts in plain text.\"]

They are typing where was Jennifer Lopez born

[\"Can add knowledge. Confirmation screens.\"]

By NYC - did you mean New York, New York?

Building an end user app, adding data is very important to get to the scale.

Let me show 2 more, then pause for questions.

IBM\u2019s Deep QA \u2013 An \u201cEntity Retrieval Engine\u201d
[Watson video from IBM \"It's going to change how we interact with information\"

What is empricism.

Pose question in natural language. An information seeking tool to make sure you get what you want. I don't have one of those.

A human standing there with carbon and water.

The computer standing there with silicon and disk.
Questions are tricky.

I think we've gone from impressed to blown away.

Reality is, being able to win a game does not mean you have conquered the language understanding task.

Worst is when WATSON crashes.

Or just starts making mistakes ... NO. \"HUMANS - Woo!\"

Push technology as hard as we possibly can.

What can we really do. It is limitless what we could apply this to.

[YouTube ref IBM and the Jeopardy Challenge]

http://www.research.ibm.com/deepqa/
Statistical retrieval of concepts/entities
Wolfram - clean

True K - inference

IBM - words/entities
Siri Virtual Personal Assistant \u2013 A \u201cDo Engine\u201d

Reasons about capabilities of external services

Conversational (spoken) interaction, with context

Personalized: learns and applies info about user
Siri, my own company, a \"do\" engine.

Does conversation - takes context. Next night. Action movies. No, closer to my house.

Reasons about capabilties of other services - basically web services.

User interface on iPod.

Number of services - analyzing source or sources to provide best answer.

Siri - now that TK and Wolfram - may send to those APIs.

http://en.wikipedia.org/wiki/Application_programming_interface
[Siri video]

\"I can show romantic place. Send it to a friend. If you know place you want, ask for reservation directly. Siri contacts OpenTable to see what is available. Movie? Avatar in 3D IMAX? Siri checks within the context of my dinner plans. See trailer, read plot, map theatres, get tickets from within Siri.
Ask open questions. \"What is happening this weekend around here?\"

Siri knows where I am because of the phone.

You interact with a dialog.

Check a different location \"How about SF?\"

Conversation about events carries over.

Lot of theatre up there - Wicked still showing.
\"Take me drunk. I'm home\"

Right thing to do is call me a cab]

9:51 [applause]

All available for free. Competitors, but I show them to you because they are great.

You can kind of see the flaws and the good parts.
Architecture connects services to \u2028what people do, where they are.
Knowledge

Data

Language
What are our best hopes of success?

Integrating many AI components into single system

Despite CALO. It is going to be a little bit of a while before they integrate.

With Siri, let the best flowers bloom
Learning from Massive Data

Web, but soon all books, music, tv/video, \u2026

We need more.
Learning from Massive Usage

Google has advantage over Siri. We went from 0 to modest number.

Where do they click. How do they use system.
The internet population is growing at enormous rate
Learning from Active Teaching & Collaborative Intelligence

Wikipedia approach vs Wolfram Alpha
Hybrid probabilistic/logical approaches

WA, TK, great on symbolic.

Google great at probablistic

Deep QA from IBM combining a bit

SIRI a bit

need to improve dramatically
Or\u2026 something completely different

Allen institute for brain science?

slicing up brain. Trying to build a map. Reverse engineer.
- What holds us back?

Software

Brittle/fragile

\u201cAnti-Moore\u2019s Law\u201d \u2013 gets slower

Ex: boot \uf0e0\uf020MS Word

In 1988, ran MS Word on 33 MHz. Windows started faster for me then.

Just a cautionary tale.

Software - anti-Moore's law
Human understanding moves slowly

Engelbart: co-evolution of technology and human understanding/adoption

In 1968 he did demo so far ahead of time

http://en.wikipedia.org/wiki/The_Mother_of_All_Demos

read/write internet with video conferencing - fade in voice

wireless

when you clicked on link - express how you wanted to see link

Ideas on ideas. In a few years it will be done.

Still taking people a long time to get their heads around it.

Idea of mouse took off

Whole idea of links ...

1998 - I was watching and internet was just taking off.

Ex: collective intelligence progress\u2026
- AI in the future: 5 Years\u2026

Everyone will have a Siri-like assistant and will rely on it increasingly for

mobile tasks

Maybe shorter time. Now acquired by Apple.

internet tasks (e.g. travel, e-commerce)

communication tasks

entertainment/attention

- AI in the future: 15 Years\u2026

Common sense knowledge models and reasoning components begin to be more feasible \u2013 systems seem \u201csmarter\u201d, more general, are less brittle, make less stupid mistakes

Contributions from the masses

Scale issues in probabilistic/logic start to resolve
- AI in the future: 25 Years\u2026

Too far ahead. That is for you guys.
- Poll Question

Robocup goal successful?

By mid-21st century, a team of fully autonomous humanoid robot soccer players shall win the soccer game, complying with the official rule of the FIFA, against the winner of the most recent World Cup. [ref]

[video Robocup 2010 Best Humanoid - YouTube video]

Adam Cheyer

adam@siri.com
Siri available for free \u2028in the Apple App Store
Q: Siri - on Blackberry?
Can't tell you.
Q: Bryce: In the search engine that let's you make suggestions. More Wikipedia based model. Does it rank answers? If \"do you know where Jennifer Lopez was born\" - OK to discard data entries/but not these?
TK has conflict resolution. If a human enters fact and it has logical conflict, it will warn you. J Lopez born in 1892 - but married in 19 something - does not compute. Will show you inference chain. Quite a system to maintain consistency.
Q2: How much thought geiven to combining human intelligence with ...
Amazon Turk. Turn humans into robots. Post tasks too hard for computers. Army of humans paid to do certain tasks. Identify face in this picture. Every time you do it you get money. Get human assistants.
In Siri, we don't have people in India. Fully automated. Some companies do use mixed solution and have some success.
Q2 followup: Could you use human intervention to improve AI.
If it is ground truth, that is data that machine learning can use -- if you get enough of it.
Connor: K Navigator video inspiring.

When can you ask Siri financial questions?

Ask, \"How can I make lots of money?\"
I'd say Siri will be able to answer that once we get a trusted source.

It will delegate.

Yahoo answers, \"Should I leave my girlfriend?\"

Asking about money - WA will give list of richest people in the world and you'll see you are not in it.
Q: Eric/Italy: In past few weeks, is there some kind of AI that is already operating that we don't recognize. Financial system beyond our control?
So, for me, AI is by definition. As soon as it works and you take it for granted, then it is not AI.

Is AI working that we don't understand - Google maps - just take it for granted.

When you talk about finance. Systems traded entirely by computer. Competing against humans and doing quite well. Blazingly fast.

Subverting the world? I don't think so.
Q: Is Siri tied to resources, conversationally. AGI will be a compilation of a bunch of narrow AI?
Probably the question. I don't have the answer. My approach to AGI is to take the best, the first, as many reasoning systems as you can and try to learn which are useful for which tasks. At a fundamental level that may be the way the brain works.

In Watson, they have reasoners.

Surprised how hard Jeopardy questions are.

Eddie Albert Camu?

Watson got it correct. How to coallesce and compound answers.

Reason about the different sources.

They way I know to go about it. Have very specific reasoners -

Open Table would be highest rated, but

All Menus would know if lasagna is on menu tonight.

May be completely different, more holistics approaches.
Q: Emiliano. What other areas are interesting?
Third company. Change.org. Help solve world's problems. A lot of what you guys are about.

AI for mediating perspectives.

Look at global problems.

Solve crime, proverty - no one person will have \"the\" idea. A collaborative process.

A system that is realtively neutral. One that can document and do conflict resolution.

Here are inconsistencies.

Transcribe, annotate across multiparty discussions.

Bring in best data.

Make it available.

Make it fit into this shape.

How do we as humanity move forward.

A computer aided assistant could be very helpful.
Neil: Let's give Adam a big hand.

5 min break

10:11

Markov NY Times article on robot teachers.

http://www.nytimes.com/2010/07/11/science/11robots.html?_r=1
10:19

Neil: Dan Barry - teacher of collaboration and communication.

[slide of geese in V formation]

Dan: So. Robot lecture. Focus on collaboration.

Last time adaptation.

The last time I talked to you it was about adaptation, but before that you need collaboration

Take home message: \"Cooperation trumps competition\"

How could that be? Cooperation is part of the definition of life.

Is it competition that wins out in the primordial soup? No, there is cooperation of molecules that get together, get organized.

Eventually you get lipids, then a cell ... life.
In order to get multicell organism again you need cooperation between cells.

Birds flying in a V is an example of cooperation within a species. Other examples: bees, ants, lions, wolves, naked mole rats (http://en.wikipedia.org/wiki/Naked_mole_rat)
Penguins, Ants, Dolphins
Interspecies cooperation = bacteria

10 times bacterial cells than body cells

100 times more DNA in bacteria than in our cells

Appendix is a way for the bacteria to survive and reboot your body's bacteria levels in the event your bacteria gets wiped out

Prisoner's Dilemma http://en.wikipedia.org/wiki/Prisoners_dilemma

You and your buddy get busted and then separated

1. If one rats the other out, person who rats gets reward, person who gets ratted on gets severly punished

2. If neither rats than both get out

3. If both rat than get slightly punished
Tit-for-Tat

http://en.wikipedia.org/wiki/Tit_for_tat
New strategy beats tit-for-tat.

Team had 60 entries, some always cooperated, some never cooperated.

Half were best players. Other half were worst. If you are willing to sacrifice. What you see in Tour de France (bicycle race).

Douglas Hofstadter http://en.wikipedia.org/wiki/Douglas_Hofstadter

suggests \"super rational\" approach - with the thought that since both players are rational and know the other player is rational they will both cooperate with eachother
80-20 rule http://en.wikipedia.org/wiki/Pareto_principle

Take it or leave it, take 50?

How many people vote yes. How many no. Some people don't vote.

How about 60/40? Have no's now. 10-20%.

Suppose person says, going to take $100.

$70/$30.

$80/$20

$90/$10 yes

Where did half say yes? At 80/20.

Your choice is to either get $20 or nothing.

In this test, the other guy doesn't exist. I was pretending. You were turning down $10. Isn't that interesting.

If we think the other guy is too greedy we will hurt ourselves to punish the other guy.

Even at $99/$1 - do you want a dollar or not? That was the question.

Almost everybody says, hell no.

Interesting.
One thing I wanted to mention. Individual winning tends to make you happy. But team winning makes you happy for a longer time.
Which is more satisfying; Playing longer on a team that ultimately loses OR play by yourself and win championship.

About half/half.

What was the best thing about space flight?

- seeing Earth

- flying, feeling like magic

- by third flight - best thing was being on team - having those friends for the rest of your life.
Robot applications.

cooperation - mutual benefit. Robots will want something out of the interaction.

Like with dogs. Other animals.

Video which demonstrates people are ready to coorperate with robots.

Tween bots. All robot does is go in a straight line. Asks, via sign, \"Help me\"

It is cute and asks for help.

\"I love this guy. He followed, trailing. Before starting to help.

Day I did this was really really hot. Tires got hot and came off. These guys figured out Sam was missing the tread. Talked about it for some time. Discovered tread was right there. Put tire back in order and sent him on his way. One of the things I found amazing. Never know how deep the level of engagement will go. This girl was amazing. Actually talking to herself\"
This robot - 30 different people help it. People bonded with this trivial robot.

There we are, helping this robot. People so happy to be helping this robot.

Also

Showed you swarm.

Cooperate to get tasks done to
Help me find applications where it is a two way street where robots and humans help each other.
EXAMPLES OF HUMAN/ROBOT COOPERATION

Use web or talk together.

1. The Matrix

2. Automatic vacuum cleaners (not roomba) - vacuum cleaners get plugged into walls and guided by humans, cooperate with humans in cleaning

3. Research robots - can collect data that the human can then analyze, humans have to set up the parameters to research

4. Blood cell robots - fighting disease or releasing oxygen inside human while human provides energy

http://en.wikipedia.org/wiki/Respirocyte

Matt: cells to transport oxygen. Blood would provide energy.

Dan: Just the sort of thing. Robots in your bloodstream. You provide energy for them to live. A symbiosis. Nice concrete example. That's the kind of thing I'm looking for.

5. Exploration robots

Emiliano: send out, report back

Dan: What have you done for robot? That's a one way street.

6. Rechargable robotic limbs that need repairs

7. Ant construction robots using swarm behavior to build structures

http://en.wikipedia.org/wiki/Ant_robotics

8. Robocars acting together

9. Intelligent homes and cars that work with humans to maintain optimal and efficient energy usage and overall mutual existence

10. Captcha http://en.wikipedia.org/wiki/CAPTCHA - helps the computer to understand \"English\" - gives humans secure login

Matt: You type in mixed up words. AI starts to recognize more complex patterns. It helps with security solution.

11. SETI at Home http://en.wikipedia.org/wiki/SETI#SETI.40home - swarm of robots

12. Davinci robot, operated by humans, gives doctors more precision and better view

13. Army bomb defuser bots, predator drones

14. Symbiotic relationship between \"walking stick\" robot and blind individual. The robot gains body heat for energy and the human gains \"sight.\"

15. Self-winding watch.

16. Human with internal robot -- both have to agree on actions.

17. Learning or helping

18. hiking helper -- we give it learning, maintenance, exposure to new environments; it gives us carrying help.
11:09 wrap up.

: Robots playing a game. Feedback - robots entertain people.

Dan: What do robots get?

Play

Dan: OK if robots get reward. What do robots want? They only want what we told them we want.

People in park not helping robot but person who invented the robot.

From the individual's point of view, they were helping the robot.

Through your whole life, parents tell you to go to medical school. You finally get a prof to give you a letter of recommendation to med school. Is that prof helping you or your parents?
Bryce: Middle sibling. An 8 yr old and a 4 yr old. The 8 yr old teaches the robot. For the 4 yr old, the robot can take on a teaching role. Learning from and doing the teaching.

Dan: Robot gets to be loved. I think that is a motivation. We are starting to get to program emotions. Middle sibling helps with sibling rivalry.
Emiliano: Marketplaces where you can outsource program tasks. Choose which tasks. Bots which can outsource to humans the improvements

\\\\\\\\

[sorry pad died]

Better sensors. Exoskeleton. Win world cup. Learn mobility/agility.

Dan: Sport robot that is the kind of thing I was looking for.

You want robot to win as a roboticist. They become more agile, develop better balance.

With sense of self, being able to move would be valuable.

Owners and players. In sport, you train players.
Maggie: Robot to assist a blind person. Robot would get energy from body heat.

Human would get better mobility.

Dan: Seeing eye dog

Maggie: Walking stick

Dan: Not sure body heat is enough. But you are touching on wearable clothing - wrapped around you, powered.

Wearable robotics that enhance your function. I like it.

Something about the seeing eye robot that has to do with independence.

Roles we are assigning are very high level functions.

Not clear robots would care about these things.

Hard exercise to REALLY effect the goals of a robot.
Candice: How do you know it has emotions and not just programmed to look like it does?

Dan: How do I know you have emotion? It's a classic philosophical argument. Does a fly have emotions? No. A cat? Absolutely. Where do you draw the line.

Will be exact same thing in pragramming.

Not sure we ever will be able to draw the line.

The programs we have now, doesn't have emotiton, draw a line in the animal kingdom when an animal has an emotion and know that and I'll let you know where the line is in understanding robot emotions.
Candice: We never developed outside our species.

Dan: Fair enough. May be very difficult for us to draw the line.

End with story.

Judge life by winning. Not about your accomplishments. About other people in your life.

If you successfully find a true love, your life will be amazing.
Seattle Special Olympics, Fred Rogers:

For the 100 yd dash, 9 contestants lined up

at sound of gun, they took off.

not long afterward a little boy, fell to the ground hurt his knee and started to cry

other 8 heard him crying, turned around. Every one ran back to him.

one little girl iwth down syndrome bent down to kiss his knee and said

\"This will make it better\"

The all linked arms and

All finished race at the same time.

Everyone stood up and cheered for a long long time.

What really matters is helping others win.

11:25 [applause]
Neil: Restart at 11:35 for Bill from IBM
Dan: How to breed chickens. Over 6 generations of picking best INDIVIDUAL layers, chickens plucked one another to death. Need to identify CLUTCH of chickens that produce the most eggs.

[Actually, \"clutch\" refers to the group of eggs: http://en.wikipedia.org/wiki/Chicken ]
/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////

AIR CL9 Reverse-Engineering the Brain

William Risk

published over 15 patents

- Cognitive Computing via Synaptronics and Supercomputing

Thank you.

Speaking on behalf of colleagues at IBM Almaden.

Setting is a very beautiful one.

This work includes contributions of work by colleagues around the world.

More recently ccollaboration with DARPA.
- Your (very) personal computer

Here is the spec sheet.

Weight - 1.5 kg

Volume 1.3L

Power ~ 20W

Processors: 10^11

Interconnects 10^15

\"Wiring\" ~150 km

Good at taking inputs and integrating into a coherent view.

We've been working on building computers to do something like this.

Make sense about how to act in noisy environment.
- Dual Objectives

Facilitate discovery

Glean from neuroscience
- Intersection of Neuroscience / Supercomputing / Nanotechnology C2S2

- Inflection Point 1: Neuroscience has matured

- Nobel Prizes in Nerve Signaling

Nice page [ref]

nobel_prizes ...

- Classic Neuroscience Textbook. 1400 pages [ref]

Erik Kandel, most often used text book (Principles of Neural Science)

- Neurons ~10,000 inputs

Dendrites, Soma, Axon

~ 10,000 outputs
On scale of 10s to 100s of microns in size.

Complex \"arbor\" of connections.
- What Neurons Do

Membrane reaches threshold, emits spike

- Communication between Neurons

via synapses

\"pre and post\" synaptic neurons

- Brain Scaling

16 m Neurons in mouse, 20 b Neurons in Human

- One way flow - 20 - 40 nM gap at synapse. From Graham Johnson.

Conductance strength - learning takes place. Memory.

Similar to electrical resistor.

How are memories stored? Lots of theories.

- Spike Timing Dependent Plasticity (STDP)

If neuron 1 spikes and then a short time later 2 spikes, there is likely an influence.

If 2 fires first, then 1 fires, unlikely there was an influence.

Increase / decrease conductance.
- Structure of Brain.

Over a century of neuroscience.

Gray Matter - 2-4 mm thick.

Zoom in, find dense packing of neurons.

White matter, just axons.

So, local and longer range connections.
- Cortical Organization: Layers

Typically 6.

Specialized for different functions.

Receiving,

Sending to other areas
- Layers in the Cat Cortex

Very careful tracing of cat cortex

Some lateral organization.

Cortex is not just in layers, but in columns. Particular functions.

Larger regions
- Regions

Visual

Audition

Motor functions

etc.
cortex is a 2 dimensional object, or maybe 2 plus epsilon.

Covers about 4 sheets of notebook paper side by side.

- Monkey Cortical Atlas

Paxinos Huang Toga 2000

[nice panning image]
What is local, what is long range, important to have map of what parts do and what is wired together.
- Interconnections Between Regions.

Felleman & Van Essen

Very tedious disections.

More modern techniques:

- Tracing White Mater Pathways

Difusion Tensor Imaging.

The way water molecules diffuse in the brain.

Tells you - roughly - which way axons are going.

Many sets of axons may account for water flow - daunting challenge to sorting challenge to sort out signals.

Try to analyze maps on supercomputers.

\"Projectome\"
- Cortical modeling

Layers / columns / regions / connections

Can create mathematical model.
- Inflection point 2 - supercomputing meets the brain.

- Simulation of 512 neurons, 1956

TESTS ON A CELL ASSEMBLY ...

- Computers have advanced. Blue Gene/P

Worth asking: what would it take to model mammalian brain?

Memory / Computation / Communication (Is there an intersection of what is required to do and whats available. )

- Mouse / Cat / Human

N 2x8x10^6 2x3.8x10^8 2x10x10^9
Computation and memory. TF to PF in computing power

TB to PB of memory
Super computers are up to that challenge.

IBM uses 150,000 CPUs to give 500 TF, 144 TB of memory.
- BlueGene Meets Brain.

On /L machine.

Both model and machine running it.

1% of human cortex.

Cat cortext /P machine. This simulation May 2009. 10 trillion synapses.
- Inevitable?

500 fastest supercomputers.

Blue is #1. Green line is SUM of their capabilities.
Full human approximately within a decade.
Machine will be huge, consume enormous amount of power. Only run models.

What about something you could run in a backpack?
- Inflection point 3: Nanotechnology meets Brain
Look at power

Rat 50 mW, Human 20 @

Space 6 cm^2 2,400 cm^2
Novel non-von Neumann Architectures are necessary.
Data from Todd Hylton

~ 10^10 synapses/cm^2 vs 10^10 intersection/cm^2 in 100 nm crossbar
~ 10^6 neurons/cm^2 vs ~5x10^8 transistors
long range axons vs ~ 30 Gbit throughput
- Electronic implementations of synapses

One wire as axon. One wire as dendrite.

If Pre fires first, Post fires second, increase conductance.

Reverse order, decrease.
Front end of the line processes.

CMOS Neuron Sheet

overlaid with Nanotech Synapse crossbar
- Ultimate Goal

do what the brain does.

Integrate the inputs. Hopefully more sophisticated than this [tiny brain in
- Research Sponsor - DARPA

SyNAPSE: Systems of Neuromorphic Adaptive Palstic Scalable Electronics

got to work with great partners
David H: Crossbar architecture. What are the technologies there. Materials basis?
One of the focuses of DARPA project - see if size/density SDP can be achieved.

Generally, physical systems where conductance can be changed on timing. Has been a key focus of the program. Several promising candidates.
David R: Each neuron has 10,000 synapses. What happens to rest of synapses?
Neuroscientists will give very detailed answer. Different mechanisms. Really fascinating. People have tried to capture that level of detail.

Another approach - more abstract. Neurons act kind of like slide - in/out - integrate and fire model.

Other part of question -are synapses independent. Better for neuroscientist. Each one may change conductance depeding on pre/post connections.
Emiliano: What happens if Penrose is right?
It is not our objective to prove it right/wrong. Not so much interested in neuroscience, but to apply to model which can be instantiated in electronics.

Architecture/ model that works.
Sam/AU: Modeled cat brain. How functional?
Movie shows that. Supercomputing sims employ a model Account for different regions of cortex. Provide input to one layer - a 2D image, like a triangle - watch how excitations are propagated. Output of these simulations is a movie.

cortex is a 2D object.

Propagation to other areas of cortex.

Challenge to abstract from huge amounts of data.

Challenge to display these movies.

Might have something with one pixel for each neuron. One screen can nearly do that.

Supercomputing center has

250 million pixel wall.

See levels of detail we could not see previously.
Q: Integrate and fire seems simplistic. Molecular coding. Not like a transistor. Also above that level - synchronization/feedback. We're not good at designing these. We building yet another backprop network or something bigger?
Constructing from what is known about the brain. How much detail do you need?

Some very faithfully describe.

Higher level uses integrate and fire - which may be too simplistic.

May just be too simplistic to capture biochem.

May be ok for function.

We are not locked into integrate and fire.

We can try different models.

Important question. Our aim is with as generalized, abstracted model as possible, find how spikes work.
Emiliano: You use simulation/ subsystem. If we are in a position to sim enough neurons, what would prevent us from replacing subsystems?
The sim effort has not only scaled up neurons, but level/complexity it is trying to describe.

Early models were of columns. Layers of connectivity.

Not white matter. Now we're getting to that.

Progress is trying to get closer to human scale, but also more realistic on communication pathways.

Understanding the brain as a network.

Can you use network theory. Long / short connections.
David R: Why interest in trying to shrink it down? Does size really matter?
Mainly the interest is to provide something portable. Serve as cognitive processor of a robot.

Lot of use cases where, for example, doctor with sensors to smell, measure patient.

Ubiquitous.

Not just size and weight, but power consumption.

If super computers keep scaling the way they do, you'll need a nuclear power plant for each one.
Thank you.

12:16 [applause]
http://en.wikipedia.org/wiki/Dharmendra_Modha
Dharmendra S Modha's Cognitive Computing Blog

\"to engineer the mind by reverse engineering the brain\"

http://www.modha.org/
Salim: Packed week. Last week of core lectures. Suck as much from faculty as possible.

Lunch time, guest speakers. Special guest - Kaufman - Fishbank at 1:30 at 583C. Talk on entrepreneurship - testing new models.

Business is new model.

What works where.

An amazing talk.

This evening Jon short session.

Try to eat lots of nutrients, vitamins.

See you in an hour.
=============== DURING-presentation notes
Can't stand the colors of Etherpad? Turn them off with Pad Options.

(Note: this changes the color for EVERYONE.)
Color card definitions:

Green - Agree

Red - Disagree

Yellow - Slow down (content) / I don't understand

Grey - Speed up we know this stuff / Move on to another topic
hands moving apart (sideways) - speak slower, speak more clearly
Link to backchan.nl for questions:

http://singularity.backchan.nl/meetings/view/165 First speaker

http://singularity.backchan.nl/meetings/view/166 Second speaker

http://singularity.backchan.nl/meetings/view/167 Third speaker
GSP10 Book List:

http://bit.ly/GSP10BookList
=============== POST-presentation notes
Instant Evaluation:

https://sites.google.com/a/singularityu.org/gsp10/feedback/week-4
https://spreadsheets.google.com/a/singularityu.org/viewform?formkey=dGZadGpDejljY0JaWnRnV2Q1SEdYRGc6MQ
Want to build an Android phone app with Google App Inventor? Start here:

http://appinventor.googlelabs.com/about/
"

http://su-etherpad.com/workshop-Aug4-Silicon-Valley-Entrepreneurship

"Wed, August 4, 9 am - 12 noon
Silicon Valley Entrepreneurship Models and Frameworks:

Mapping the Territory

FEE

B.583c Atrium
David Rose

Tom Byers

Slides: http://bit.ly/ahXIDZ
Aaron Kemmer

Alaeddine Mokri

x Alex

Alison Lewis

Anders Hvid

Ankur Jain
x Bill Bing

Chiara Giovenzana

Chiara Turelli

Connor Dickie

David Roberts

-

David Wyler

Dhaval Chadha

Dmitriy Tseliakhovich

Emem Andrew

x Erez Livneh
x Eric Ezechieli

Erika Anderson

Eugenie Rives

Everson Lopes

Francesco Galietti

-

Gary Gautier

Hind Ahmed

Jason Dunn

Javier Mares

John Graves
Jorge Fernandez

Juan Martinez-Barea

Julian Ugarte

Kausar Samli

Luca Escoffier

-

Mercy Njima

Michael Jensen

Miguel Oroz

Nolene Naidu

Raycho Raychev
Robertq Denning

Ronen Amit

Santiago Bilinkis

Sasah Grujicic

Steve Cronin

-

Tigist Ashenaffi

x Tony Lyu

Tyler Kratz

Vincent Daranyi

Zain Jaffer
45 Students
David Rose:

Three very different subjects.

- The Essence of Entrepreneurship
Tina Seelig - [most]

Do earlier.

More next 2 days about entrepreneurship.

How many are? [100%]
What the hell is entrepreneurship.

Tom Beyers - runs e program at Stanford

Written book.

Silicon Valley eco-system.

How firms form here.

California

Thursday - detail of finance.

Tom will touch today

VCs touched last night.

Angel

Super Angel

Friends and family

What you are looking for.

Elevator pitch

4 hour work week author.
Tim

There are those of us who have 400 hour workweek.

Opposite end of cages in a zoo.

Look at Tim - nice work if you can get it.

Tim is a good speaker.

Interesting stuff to say.
This morning. What is the essence of being an e

What does it mean to you?

You are what?
Eric: Start with one idea. Transform idea into results.
Bing: See opportunity where others see rick.
: Willing to take risk

: Good salesman -

Not in official definition - does require other people to buy

product or vision.
What does the dictionary say?

We will deconstruct definition.

A person who -

stop right there. A person.

Companies have intrapreneurs.

Comes down to a very personal thing.

Each of you is a person

Very hard to codify.

Hear from Tina how to jumpstart innovation.

From corporate basis - from Tim

All progress depends on the unreasonable man.
\"organizes and operates\"

You start it.

Could be someone else's idea.

Not innovation. Organizing.

Difference between inventor and entrepreneur.

Lots of people, particularly academics - not e. Ideas will never get out of lab.

Earlier in the stage - how do you match idea with someone who can make something out of it.

If you just organize and run away - a founder.

To be an entrepreneur you need to override the market,

actually taking action. Run a business.

Relatively few entrepreneurs start and run through.

Rare to continue to run for a long time.

Alex: Bought themselves a job.
John Kastner -

Princeton Review.

Not an entrepreneur - was in incubator - second company starting makes you an e.

Latest doing really cool stuff. Taking education - going on-line

Outsourced, on-line for high end.

2tor

http://2tor.com/

Providing on-line systems for universities.

Raised $20 m.

Typically, the way e works. Vs Small Bus person.

typically, like peanuts - once you start one you are in the flow.

People in this class have started 30.

Several have done 2 or more.

Trick - having started it, to get it going.

Like car. Start it, drive it, then hire professional chauffeur.
\"a business or businesses\"

Social - concept of social entrepreneurship - essentially a business.

Overrides the market.

Play by different rules.

Create a business. Or businesses. You tend to be serial e.

\"taking on greater than normal\"

Any e here not \"greater than normal\" going outside the system

lack of sleep

\"financial risks to do so\"

Taking on greater than normal financial risk to start organize and operate.

Business, starting, personal.

That's what you guys are.
- Adam Smith

Talk now about that in overall context

Wealth of Nations

\"By pursuing his own interest [the individual] frequently promotes that of the society more effectually than when he intends to promote it.\"
I'm a dyed in the wool liberal. Also a good capitalist.

People who do social things look down nose at business people.

Absolutely not antithetical.

Question of sustainability - keynote of last 3 days

Garvin - not just economic, but social sustainability.

Business - if social, better work economically - or it will fail.
You can do something rational that makes sense.

BUT if you say, \"I'm going to do something great for the world\"

You won't survive.
- Joseph Schumpeter

Father of entrepreneurship and innovation

Coined the term \"creative destruction\"

across markets and industries

Only by destroying the old stuff do you create new stuff.

Destruction and renewal.

Jingle networks. 411 books.

Directory assistance calls.

People have phones.

You need to find someone else's number

First operators.

Then phone books.

Then let operator check books.

People would do that.

Lots of people calling, so phone companies started calling.

Then mobile phones came along.

From mobile phone - what changed?

When first mobile phones came out -

: You don't want to write it down

: You don't have access to phone book

A choice.

Everybody had phonebooks.

Directory assistance is a luxury -> necessity.

You don't have option to look in phonebook.

Call 411 = but now phone company has monopoly.

You need a number or you can't call anybody.

They have got you over a barrel.

Have to have it.

They have unlimited margin.

Directory assistance - charged $2 per call.

Very large percentage of revenues

98% margin

So, how to destroy this business?

Give same answer for free (with ad)

Directory assistance via computer -

almost all voice recognition.

Technology shrunk the cost.

Put at end of free 800 number. Play ad.

Get paid for the ad, and since I know what you

are looking for, I can give you a targeted ad.

Redirect you -

UNOs is having a special - press * to be connected.

Very important - connect at time and place of purchase.

For customer - pay $2. OR FREE with ad.

They took 5 - 10 % within 6 months of launch.

They cut out $1.90 of the whole world.

Changed whole market.

Still legacy market.

But that is the kind of creative destruction we are talking about here.
- Entrepreneurial skills

Things you need, or need to get

People

You can create your own magazine, but

if you are going to have any scale - you need to have people

skills. Some so brilliant, can be total assholes. Others

need to be able to convince

Management
Production

Know how to do widget. Difficult to organize and operate if

you don't know what you are doing.
Operations

This takes a lot of work. Tiny fiddling pieces. Many we don't

like to do. Essential part of life.
Money (Financial Management)
Sales

Yeah. Sales are critical. More and more pitches with IP, widget,

but can't convince anybody to do this.

To show how critical sales are -

Angelsoft

Collaboration, etc.

We went to Angel groups - buy-in basis.

$100 per plan. 100 plans, pay $10K.

So we'll automatically bill you.

\"PAY for software\"

So I said, OK. What will we do.

When entrepreneurs who want to get in front of you,

we'll change e.

\"CHARGE e?\"

OK, you won't pay. You won't let us charge the guys who would pay.

What about sponsors?

When e apply, we'll charge the sponsors.

\"CHARGE them! The buy our dinners.\"

Will it save money. Something you want and need.

We've got that done.

So ...

OK fine. We've got a bigger fish we are frying.

We'll give you this for free.

\"We can't take it for free\"

You won't???

Does it do exactly what you want?

Why won't you take it free.

So, you won't make money.

Angel investors are psychotic people.

Crazy squared.

OK. Sales is it. 4 person team.

Take this software, PLEASE.

Chase them into bathrooms, begging them to take it free.

Sales is really critical.
Financing.

How do you get cash in the first place.

Will talk in detail Thursday.



Share with your friends:
1   ...   9   10   11   12   13   14   15   16   ...   57




The database is protected by copyright ©essaydocs.org 2020
send message

    Main page