Go and AI

[Charles Matthews]

The announcement that go will be a solved problem, in Artificial Intelligence terms, by about 2016,

I know all the words but have no idea what this means. I'm guessing it could just mean they may know the true size of komi, not that a machine will beat all humans????

Also, if go is solved by 2016, shouldn't we expect chess to be solved tomorrow?

I'm sure what he meant was at the level of Deep Blue and chess. The wording is mine, of course. Too much is being read into it verbally (a 19x19 failing, indeed). The only clear prediction made is of dramatic progress in level.

The thing is that Demis does actually know something about traditional games.

[Codexus]

I think in this context it was meant as "it can play the game as well as human champions". At least this the claim I have been hearing lately: that within a few years, computers will beat humans at go.

The new piece of the puzzle is the so called "deep learning". From what I have seen so far that seems to be the same as good old neural networks but apparently now they work much better. After training them with a large data set of human games, they are apparently not bad at predicting the likely next move.

That alone does not make a go AI that is stronger than current programs. But the idea would be to combine that with the Monte-Carlo Tree Search algorithm used in the current generation of go AIs. So basically instead of playing random games, the MCTS would use the neural network to get its candidate moves. Which could possibly result in a pro level go AI according to some people. (that seems a bit of a leap of faith to me but it's possible)

In that sense, the way to build a champion level go AI has been almost solved from a research point of view, and all that's left is some software engineering to actually combine the parts

I'm not entirely convinced this is the end of the story...

[Charles Matthews]

I'm not entirely convinced this is the end of the story...

Nor I. Odd that no one's reaction in this thread is "good for the game".

[John Fairbairn]

Odd that no one's reaction in this thread is "good for the game".

But no-one appears to have said "bad for the game" either. My guess is that people in the games world have already absorbed the chess experience and would just see a go computer as another tool - in other words, be rather neutral.

That said, there is an abiding problem with chess computers in that they can't explain (or even actually) prove to humans why one move is better than another. Conceivably, neural networks might shed some insight for a human, but the present paper talks about Monte Carlo to shore up the bulwarks, so this research doesn't seem to be going in a promising direction for providing illumination. Hence, it probably isn't specially good for the game except, like Deep Blue, for a passing moment of PR.

[Charles Matthews]

Hence, it probably isn't specially good for the game except, like Deep Blue, for a passing moment of PR.

With the difference that it is rare to find a person who hasn't heard of chess; while the average person hasn't heard of go (in the West). The "Google takes seriously" cachet is quite something.

[Krama]

I think in this context it was meant as "it can play the game as well as human champions". At least this the claim I have been hearing lately: that within a few years, computers will beat humans at go.

The new piece of the puzzle is the so called "deep learning". From what I have seen so far that seems to be the same as good old neural networks but apparently now they work much better. After training them with a large data set of human games, they are apparently not bad at predicting the likely next move.

That alone does not make a go AI that is stronger than current programs. But the idea would be to combine that with the Monte-Carlo Tree Search algorithm used in the current generation of go AIs. So basically instead of playing random games, the MCTS would use the neural network to get its candidate moves. Which could possibly result in a pro level go AI according to some people. (that seems a bit of a leap of faith to me but it's possible)

In that sense, the way to build a champion level go AI has been almost solved from a research point of view, and all that's left is some software engineering to actually combine the parts

I'm not entirely convinced this is the end of the story...

This is what I was also thinking about.

Since we can say the network can guess the correct move (one played by pros) 55% of the time. Then why not order the network to give us let's say 5 moves. If the correct move is found 55% of the time in 5 moves there is a 95% chance that one of those moves would be played by a pro. With 10 moves selected it goes into 99.75% but it would probably be an overkill and slow the whole process down.

So asking a network to give us top 5 moves, we can be pretty certain that we get a correct move. However if we repeat this process with let's say 125 turns (if we take that an average game lasts 250 moves) then something bad happens. In 6 turns out of those 125 you get all wrong moves. However if you increase the number of moves you get from the network to 10 then you can pretty much be certain that you will get at least one of those 10 correct through the whole game. However I still have so much things on my mind that can prove this method not so good but I just can't bother to write another walls of text. Maybe in the next couple of days I find some time to write some other ideas on how to improve go AI.

[Uberdude]

They deliver: http://www.lifein19x19.com/forum/viewto ... 23#p198123

Time to pay up palapiku?

[Mike Novack]

Since we can say the network can guess the correct move (one played by pros) 55% of the time. Then why not order the network to give us let's say 5 moves. If the correct move is found 55% of the time in 5 moves there is a 95% chance that one of those moves would be played by a pro. With 10 moves selected it goes into 99.75% but it would probably be an overkill and slow the whole process down...................

You are missing something (misunderstanding "right move" and "wrong move")

There (usually) isn't ONE move made by pros (for most of the moves in a game). A neural net is trained on the records of actual games played by pros. When you say 55% that's the move made in that situation by that pro. But when it returns some different move, that's the wrong prediction (of what the pro did in that game) but NOT necessarily the wrong move. Not necessarily a BAD move. A little thought and you would realize that on at least move in each game between top pros a mistake was made (somebody lost, didn't they). So even when the neural net returned the "right move" (the move the pro made) that might actually be a "wrong move" (in terms of the game).

So a modification that returned five moves (that set of five moves) might have a high probability for the set to contain the best move. MCTS could be applied to the five.

[hyperpape]

No need to be condescending.

[pookpooi]

I feel heartwarming, because you know what, Demis Hassabis, your go disciple, is honest about his word, he deliver it at the beginning of 2016, and I hope he'll close the case (by beating the current champ, not Lee Sedol) in 2016 as well.

[Charles Matthews]

I feel heartwarming, because you know what, Demis Hassabis, your go disciple, is honest about his word, he deliver it at the beginning of 2016, and I hope he'll close the case (by beating the current champ, not Lee Sedol) in 2016 as well.

Thanks for the kind words.

If AlphaGo can establish some sort of claim to be in the 4p to 7p bracket, then the whole business of computer go is clearly changed. One victory against Lee Sedol that wasn't decided by a blunder might do that.

So far I think we know the program has "talent", and is worth watching from a spectator's point of view.

[pookpooi]

I feel heartwarming, because you know what, Demis Hassabis, your go disciple, is honest about his word, he deliver it at the beginning of 2016, and I hope he'll close the case (by beating the current champ, not Lee Sedol) in 2016 as well.

Thanks for the kind words.

If AlphaGo can establish some sort of claim to be in the 4p to 7p bracket, then the whole business of computer go is clearly changed. One victory against Lee Sedol that wasn't decided by a blunder might do that.

So far I think we know the program has "talent", and is worth watching from a spectator's point of view.

Actually, I'm glad that I didn't read this topic back in 2015, cause I might reply with something like 'I don't believe you'

Some food for thought, Google also buy another frontier company in making robot with legs, Boston Dynamic
and what if you combine the best software from Deepmind and the best hardware from Boston Dynamic and let it challenge human in the game of humanity... yes, I mean football/soccer! The idea and condition are very rough right now but you know it's gonna happen at some point, hope it's in my lifetime though.

[Charles Matthews]

Some food for thought, Google also buy another frontier company in making robot with legs, Boston Dynamicand what if you combine the best software from Deepmind and the best hardware from Boston Dynamic and let it challenge human in the game of humanity... yes, I mean football/soccer!

I guess "robot with wheels or wings" is more à propos. DeepMind will turn to 3D games, according to Demis Hassabis in a video interview. So, cars that drive themselves, drones ... all that mundane stuff.

[pookpooi]

Some food for thought, Google also buy another frontier company in making robot with legs, Boston Dynamicand what if you combine the best software from Deepmind and the best hardware from Boston Dynamic and let it challenge human in the game of humanity... yes, I mean football/soccer!

I guess "robot with wheels or wings" is more à propos. DeepMind will turn to 3D games, according to Demis Hassabis in a video interview. So, cars that drive themselves, drones ... all that mundane stuff.

By the way, do you know how good Hassabis is in go? He states that he's not a strong go player, but a very strong chess player though. I'm guessing he's in single digit kyu?

[gowan]

Odd that no one's reaction in this thread is "good for the game".

But no-one appears to have said "bad for the game" either. My guess is that people in the games world have already absorbed the chess experience and would just see a go computer as another tool - in other words, be rather neutral.

That said, there is an abiding problem with chess computers in that they can't explain (or even actually) prove to humans why one move is better than another. Conceivably, neural networks might shed some insight for a human, but the present paper talks about Monte Carlo to shore up the bulwarks, so this research doesn't seem to be going in a promising direction for providing illumination. Hence, it probably isn't specially good for the game except, like Deep Blue, for a passing moment of PR.

Based on what I know about AI neural networks I doubt that this program can explain in conceptual terms why the moves it chooses are good.