Foresight Exchange – Selected Claims + Commentary
The Foresight Exchange is a spiffy site. People get together and bet fake money that various futurist events will occur. The idea is that the environment of a prediction market will let you see the aggregate views of a group of betters in a way more insightful than just voting. Bets introduce the dimension of magnitude of confidence. In theory, the urge to win also causes people to think things through more carefully than they would if they had to answer off the cuff.
Naturally, when visiting the site, I go right to the "Computing Theory/AI section". You can see the various lists of claims here. To make your life easier, I'm going to post the interesting AI-related bets here along with their "price plots" (probability graphs) over time.
Ticker symbol: GoCh
The Claim:
A machine will be the best Go player in the world, sometime before the end of 2020 (even if subsequently eclipsed by human players). The evidence and playing conditions are the same as for the Machine Chess Champion Claim, with time limits etc. adjusted for the playing customs of human high-level Go competition.
Price plot:
People are lukewarm on this one. Available computing power per dollar is predicted to increase 100 times over by 2020. I would think this would be enough to brute-force the problem, but it's hard to tell for sure. True brute force is not possible - there are far more possible Go games than chess games - but a Go AI could go pretty far simply by using a large set of heuristics and training set data consisting of games played by the greatest masters. By 2020, I think that Go AI should be good to go.
Next one...
Ticker symbol: mdvw
The Claim:
Before 2025:
1. A mindviewing device can detect the processing of the human mind.
2. The device helps us understand most of the sentence of mind in brain.
3. The device is approved by most scientists.
What 2 is trying to say is that you could use the device to read out most of a sentence that some guy has in his head, without any corpus limits or special training.
Price plot:
Most people obviously don't believe this one will happen. I think they are wrong. We can already make a video of what a cat is seeing by implanting probes in kitty's brain. The main thing holding back human brain-computer interfacing research are ethics regulations against most human testing. New testing noninvasive testing methods, like transcranial magnetic stimulation (TMS) are circumventing these ethics limits. Ian C. also posted this interesting link yesterday - about a guy trying to create a brain chip that can recreate thoughts. Basic NT (just 'nanotechnology', not even molecular nanotechnology) will allow million-unit-and-up brain probe arrays. Words in our brain are stored in something called the phonological loop, and even if working memory is really defined as interneural connection weights across large portions of the brain and is not strictly localized, I still believe that we'll be able to read sentences in people's mind by then.
Next!
Ticker symbol: Robbie
The Claim:
By the end of December 31, 2035, a humanly mobile robot will be in actual use, as a worker in some business use, consumer service, military use, civil service, or scientific capacity. The robot must be generally human-like or humanoid in shape (see "Background", below), and must have a very human-like competency of motion, along with the ability to distinguish between adult humans, animals, and young children, in circumstances in which the "object" in question is in plain view, *and* in which there is no costuming, trickery, or other mitigating circumstance that might cause even a human to be mistaken.
Price plot:
This one is pretty certain. 2035 looks conservative. 2025 seems more likely. It's true that all the Asimo demos are carefully choreographed, plus it recently managed to face plant while walking down stairs, so we aren't there yet, but it's only a matter of time. If the human motor neurocircuitry is too hard to recreate independently, we can just copy it. The computers and scanners to achieve that will be here in the mid-20s if the rate of progress continues as it has. Such a robot would have a reprogrammable motor lobe, allowing it to act like a ballerina one second and a martial artist the next.
Of course, the introduction of a non-human robot with the full human range of motion would go a long way towards eliminating most conventional labor. They could also build more copies of one another. If that last sentence made you slightly afraid, then you probably have an Adversarial Attitude towards robotics and AI, which tends to get in the way of thinking reasonably about it. Get rid of it and you'll be happy you did.
All done for now. There are plenty of interesting claims on other topics, I suggest you check them out. Have a nice weekend!
April 7th, 2007 - 01:26
“Available computing power per dollar is predicted to increase 100 times over by 2020.” – the factor is more likely to be between 5000 and 10,000.
Since I end up figuring these a few times a year, I went ahead and created a little calculator:
http://hyper.to/blog/cps.html
It has two fields for year, and calculates the compute powers and ratio. It also compares the compute powers to that required to compete with (or simulate) a human brain.
April 7th, 2007 - 09:05
I was being conservative and going with the 2-year doubling time, as Moore originally stated. But yeah, 5,000-10,000, if not BILLIONS TIMES MORE, due to molecular manufacturing, is more likely.
Thanks for coding up this little program, Miron.
April 8th, 2007 - 09:15
I’d like Foresight Exchange more if it allowed the betting of real money. One could earn a small fortune that way.
April 9th, 2007 - 00:45
Kaj, I don’t think that would be legal in the US due to silly anti-gambling laws. I definitely think it would be a good thing. For one it would make the predictions much more accurate, since it would give a more concrete incentive to the participants. This would improve their research and reduce, somewhat, their emotional bias.
Michael – agreed on the potential for billions… but even without additional acceleration, we’re in for some very interesting times in about 10 years. Which is suddenly looking like a pretty short time.
April 9th, 2007 - 22:53
If that last sentence made you slightly afraid, then you probably have an Adversarial Attitude towards robotics and AI, which tends to get in the way of thinking reasonably about it.
I’m not so much concerned about the Adversarial Attitude toward the bots as toward their owners. Then again, it’s the first sentence that’s the important one; the next one just exacerbates the problem.
Sure, anthropomorphizing AI is unreasonable, but anthropomorphizing humans is much better justified.
April 10th, 2007 - 10:04
Actually, given that global population growth rates indicate that between 2040 & 2050, the average population of the human race will surpass 50 years, with more seniors above the age of 65 than there are people below the age of 25; if we DON’T have anthropophenomic robots in general industry w/ low-grade AGI, we will discover that our economy will have little choice but to collapse in upon itself. (Anthropophenomic for those who pidgin even less latin/greek than I do means “Human appearing”)
That is to say, humanoid robots shouldn’t be something to fear. They should be viewed as the potential salvation of the modern human experience.
May 15th, 2012 - 19:29
Nice post at Foresight Exchange – Selected Claims + Commentary Accelerating Future. I was checking continuously this blog and I’m impressed! Extremely useful info specially the last part :) I care for such information a lot. I was seeking this certain information for a very long time. Thank you and good luck.