You're describing Opera before they for some fucked up reason decided to throw their code away and rewrite it as a Chromium skin. I'm still weeping over it.
Why would you want to look at all possible moves? Humans don't do this, why would it be necessary for computers?
Yes you do. You just use intuition to skip over moves that might not be worth your time, but you still consider them. AlphaGo does something similar with a neural network before brute-forcing into good possible moves.
Still, even if you don't want to consider 10^700 possible game trees on a clean Go board, the problem is still intractable. Go has, in average, 250 possible value moves to consider after each stone is placed. Chess has around 30.
Go is rather simple compared to other problems like image recognition....
No, not really. We've had relatively strong image recognition algorithms for a good while now, and i'm not talking just about Google Image Search. Image sensors have been used for a long while in industrial automation settings, from anything for measuring to actively identify features in production lines. As a problem is way more accessible than Go is.
The real question is: What is intuition? Is it something computable or not? If it is only some kind of statistical inference, then no wonder we are good at it: we have an inference engine which structure has been optimized by million years of evolution, and fed with bazillions of samples since our birth.
Agreed. One could argue that the way AlphaGo picks up moves (adaptive neural network) is "intuitive"; we don't know really what drives after some training. A Google engineer today cannot really tell you why AlphaGo favored some moves over others.
He's likely to be remembered as the last human being to beat a Go AI on tournaments.
Move 78, in particular, was so good that his partners and commentators in China have already called it "the hand of God", but it really was one of those things which happens once in a blue moon, even for a player like Sedol.
The F-35 at 1.3T is replacing 3 models of aircraft. It's current LRIP fly away cost is _already_ cheaper than a new F-15, let alone an F-22 (when it was in production).
No, it is intended to replace three models. I'm still waiting on any explanation (official or not) on how a F-35s is expected to replace the A-10 on CAS duties. Hell, the US might be better off just buying a fleet of Super Tucanos from Embraer and writing it off as losses on the total program cost.
You likely mean the F-104. That aircraft was basically a rocket-powered dart - it had such a poor glide ratio that if you experienced a flameout below 15,000ft the standard procedure was to eject and ditch the plane.
This. The F-35 program has eaten ofer $1.3 *trillion* US dollars by now. It is insane; you could use that money to triple the current F-22 fleet (a vast superior, but also more expensive, aircraft) the US has and still have left over cash to fund its entire program all over again.
At this point is hard to justify the F-35s existence. It is cheaper to build and fly than the F-22, yes, but so much money has been pour on the project that those savings will never offset its cost. Same as with the possibility of exporting it; the US won't make a dime on it. On top of that is a modern fifth generation aircraft that does pretty much everything worse than the model it is replacing... and the ones used by other nations.
There's much more to Poker than just computing probabilities. Someone wrote a much better explanation on the/. story for the first AlphaGo win against Sedol but, in a nutshell, how you play on Poker plays a substantial role. If you play those perfect probabilities alone you'll loose because after a while no one will bet you against them.
What you mention happens on Poker as well. The rules are simpler, yes, but those nuances are still present - for example, you have to be careful when you bet because you're both guessing what the rest of the players are up to and potentially revealing your intentions in the process as well. Bluffing is not easy to model on AI.
Poker is often cited as an example of "imperfect information" game, where odd calculation alone will not help you win. There's already a fair amount of research on it.
Search space, basically, and the amount of moves you have to inspect before selecting a best one. Chess has about 10^120 possible moves, but you can reduce this using opening books and heuristics to a sensible number which still lets you pick very strong moves. At that point it is just a matter of throwing CPU power at the problem.
Go is a completely different beast though. A "small" 13x13 board has 10^170 valid moves, and the count for a 21x21 board is well over 10^210. So for even small, beginner-level sized boards no amount of CPU power, now or in the future, is bound to help you solve the problem. Go is interesting because every engine out there uses some form of adaptative AI - AlphaGo uses a machine learning neural net which had to be trained like a human would, but with over 30 million recorded moves.
This is nothing like the "regular" AI used on modern chess engines - those are useless for the essentially infinite game tree possibilities like the ones presented in Go. AlpaGo decides moves using a machine learning neural network and then selects the best one using classic heuristics.
Slashdot Mirror
You're describing Opera before they for some fucked up reason decided to throw their code away and rewrite it as a Chromium skin. I'm still weeping over it.
Why would you want to look at all possible moves? Humans don't do this, why would it be necessary for computers?
Yes you do. You just use intuition to skip over moves that might not be worth your time, but you still consider them. AlphaGo does something similar with a neural network before brute-forcing into good possible moves.
Still, even if you don't want to consider 10^700 possible game trees on a clean Go board, the problem is still intractable. Go has, in average, 250 possible value moves to consider after each stone is placed. Chess has around 30.
Go is rather simple compared to other problems like image recognition....
No, not really. We've had relatively strong image recognition algorithms for a good while now, and i'm not talking just about Google Image Search. Image sensors have been used for a long while in industrial automation settings, from anything for measuring to actively identify features in production lines. As a problem is way more accessible than Go is.
The real question is: What is intuition? Is it something computable or not? If it is only some kind of statistical inference, then no wonder we are good at it: we have an inference engine which structure has been optimized by million years of evolution, and fed with bazillions of samples since our birth.
Agreed. One could argue that the way AlphaGo picks up moves (adaptive neural network) is "intuitive"; we don't know really what drives after some training. A Google engineer today cannot really tell you why AlphaGo favored some moves over others.
I was wondering about that too. The funny thing is, because of how AlphaGo learns and plays moves, Google engineers cannot really tell either.
Um, no. Obvious differences asides this is like being angry at a juicer because it writes better music than you.
He's likely to be remembered as the last human being to beat a Go AI on tournaments.
Move 78, in particular, was so good that his partners and commentators in China have already called it "the hand of God", but it really was one of those things which happens once in a blue moon, even for a player like Sedol.
I can totally see neural network AIs like AlphaGo used in the HFT world. In fact i'd be surprised if there's no one investigating this as we speak.
Didn't think i'd see this happening for a long time. Wonder whats next now...
It runs in a virtual machine and my Oracle rep tells me those are bulletproof!
It is. In fact the F-35 is intended to replace four aircraft within the US military: F-16, A-10, F/A-18 and the AV-8B. Get your facts checked.
Thank you. The overall reaction (from the commentators and Lee Sedol himself) is priceless.
The F-35 at 1.3T is replacing 3 models of aircraft. It's current LRIP fly away cost is _already_ cheaper than a new F-15, let alone an F-22 (when it was in production).
No, it is intended to replace three models. I'm still waiting on any explanation (official or not) on how a F-35s is expected to replace the A-10 on CAS duties. Hell, the US might be better off just buying a fleet of Super Tucanos from Embraer and writing it off as losses on the total program cost.
You likely mean the F-104. That aircraft was basically a rocket-powered dart - it had such a poor glide ratio that if you experienced a flameout below 15,000ft the standard procedure was to eject and ditch the plane.
This. The F-35 program has eaten ofer $1.3 *trillion* US dollars by now. It is insane; you could use that money to triple the current F-22 fleet (a vast superior, but also more expensive, aircraft) the US has and still have left over cash to fund its entire program all over again.
At this point is hard to justify the F-35s existence. It is cheaper to build and fly than the F-22, yes, but so much money has been pour on the project that those savings will never offset its cost. Same as with the possibility of exporting it; the US won't make a dime on it. On top of that is a modern fifth generation aircraft that does pretty much everything worse than the model it is replacing... and the ones used by other nations.
They're not. AlphaGo is basically an AI making very educated guesses and then calculating moves.
Food for though: couldn't you argue the same about a professional Go player?
No. Not by a long shot, i might add. There's extensive studies on chess openings and endgames, but doesn't even cover a fraction of it.
There's much more to Poker than just computing probabilities. Someone wrote a much better explanation on the /. story for the first AlphaGo win against Sedol but, in a nutshell, how you play on Poker plays a substantial role. If you play those perfect probabilities alone you'll loose because after a while no one will bet you against them.
It is impossible to calculate all possible Go moves, even for Google. In fact, it is impossible to do so for even a infinitesimal fraction of them.
Good call, thanks. 19x19 is still over 10^700 valid game trees to consider though...
FIX: 21x21 is 10^976 valid moves. Mistaken by an inch there :)
What you mention happens on Poker as well. The rules are simpler, yes, but those nuances are still present - for example, you have to be careful when you bet because you're both guessing what the rest of the players are up to and potentially revealing your intentions in the process as well. Bluffing is not easy to model on AI.
Poker is often cited as an example of "imperfect information" game, where odd calculation alone will not help you win. There's already a fair amount of research on it.
Search space, basically, and the amount of moves you have to inspect before selecting a best one. Chess has about 10^120 possible moves, but you can reduce this using opening books and heuristics to a sensible number which still lets you pick very strong moves. At that point it is just a matter of throwing CPU power at the problem.
Go is a completely different beast though. A "small" 13x13 board has 10^170 valid moves, and the count for a 21x21 board is well over 10^210. So for even small, beginner-level sized boards no amount of CPU power, now or in the future, is bound to help you solve the problem. Go is interesting because every engine out there uses some form of adaptative AI - AlphaGo uses a machine learning neural net which had to be trained like a human would, but with over 30 million recorded moves.
This is nothing like the "regular" AI used on modern chess engines - those are useless for the essentially infinite game tree possibilities like the ones presented in Go. AlpaGo decides moves using a machine learning neural network and then selects the best one using classic heuristics.