Considering he's beaten Lee Sedol and Gu Li, who knows how this will play out.
But no doubt he'll talk to Sedol about his take away from playing AlphaGo, especially since part of playing against any computer is reverse engineering its decision tree.
Gu Li and Ke Jie played each other (in a jubango -- a ten game series -- in 2014: an excellent series of games!). AlphaGo did not play Gu Li unfortunately.
I'm guessing that AlphaGo has been doing more learning. Studying Sedol's games is probably a good platform but certainly not a plan to win - Jie needs to go in expecting the unexpected. Just look at how different the games were between Hui and Sedol.
To answer your question more directly: giving AlphaGo more hardware would only give it more CPU cycles to work with (faster decisions, not necessarily good ones). The true improvement will likely come from the very same source that humans get it: practice and mastery.
Yes, the online hardware used during the game, vs. the offline hardware used for training the neural network, is mainly there to run a ton of monte carlo tree search (MCTS) iterations. MCTS tends to produce better estimates as you throw more compute power at it. In perfect-information, deterministic games, it theoretically converges on optimal play given infinite compute power, although in games with very large state spaces in practice that guarantee isn't too useful on its own, and the heuristic guidance provided by the NN is crucial.
In the paper they state that they tried giving it more power than the version that played Sedol, but found that it had rapidly decreasing marginal value, so they decided on a level that was pretty close to optimal.
I assume Alpha Go has been trained using a large number of games played by Go masters so the games it played itself probably won't provide much more information, unless it can learn its own weaknesses from those games.
They said it was trained with the games of 5 dan and stronger players from KGS, plus self playing. No professional games and no training with the games of a specific player, because there so few of them (60-70 per year).
Although this is not the main reason, there is also a matter of intellectual property of Go game records. The official position of Korea Baduk Association is that Go game records are copyrightable.
On the other hand, it is not recognized as such in Korea, let alone in other countries. There has been multiple attempts to change the Korean copyright law to explicitly list Go game records as an example of work that can be copyrighted. All such attempts failed so far.
Google probably will have many times the ASICs compared to what it had against Lee Sedol, and there have been lots of new tricks in the past few months with convolutional networks that increased learning speed and accuracy of the learned model. It would be a huge win for Ke Jie even if he could win only 1 game.
An interesting thing about that rating site is that it uses Whole-History Rating, a system developed by Remi Coulom which takes the entire match history into account.
After the Sedol vs. AlphaGo games, you could reasonably ask if Sedol was on top of his game, or if his losses were in part a product of a declining rating of his own. WHR effectively considers this possibility, and the similar possibility in all matches - you could say the long-term impact of a match doesn't stabilize after both participants have played a few more games.
AlphaGo has been steadily climbing in this rating. Sedol doesn't appear to be in decline, in fact he had an 8-win streak right after the AlphaGo matches (ended by Gu Li just recently).
This will be interesting. After it's first game win against Lee Sedol, Ke was very confident he could beat Alphago, but the later games against Lee revealed further unexpected strengths in Alphago's capabilities. I think everyone, including Ke Jie know that eventually Alphago will be unbeatable so I think it's important and useful to have a match soon when Alphago's game is still improving but potentially close to human level. I think a match at this stage would be much more revealing about how AI systems like this develop and improve than a game say in a few years time when the computer player just steamrolls to victory every game.
19 comments
[ 14.6 ms ] story [ 57.3 ms ] threadBut no doubt he'll talk to Sedol about his take away from playing AlphaGo, especially since part of playing against any computer is reverse engineering its decision tree.
Exciting times, I can't wait. I guess the later the game will be, the less chance Ke Jie has.
To answer your question more directly: giving AlphaGo more hardware would only give it more CPU cycles to work with (faster decisions, not necessarily good ones). The true improvement will likely come from the very same source that humans get it: practice and mastery.
On the other hand, it is not recognized as such in Korea, let alone in other countries. There has been multiple attempts to change the Korean copyright law to explicitly list Go game records as an example of work that can be copyrighted. All such attempts failed so far.
Who owns the copyright?
An interesting thing about that rating site is that it uses Whole-History Rating, a system developed by Remi Coulom which takes the entire match history into account.
After the Sedol vs. AlphaGo games, you could reasonably ask if Sedol was on top of his game, or if his losses were in part a product of a declining rating of his own. WHR effectively considers this possibility, and the similar possibility in all matches - you could say the long-term impact of a match doesn't stabilize after both participants have played a few more games.
AlphaGo has been steadily climbing in this rating. Sedol doesn't appear to be in decline, in fact he had an 8-win streak right after the AlphaGo matches (ended by Gu Li just recently).
"contrary to internet rumours, we've not decided yet what to do next with #AlphaGo, once we have, there will be an official announcement here"
https://twitter.com/demishassabis/status/739832323160563712