Lee Sedol versus AlphaGo
I just wanted to get out there my prediction for the upcoming match between one of the best go players in the world and AlphaGo, the computer who defeated a professional grade go player for the first time. If you don't know anything about go - go read about it here or here. Let me just put it out there; I think it's going to be a win for Alpha go, 4-1 or maybe even 5-0 and here’s why; go computing has made such a huge leap, were not going to see the slow build up the way we did in chess. In chess terms go computing just went from somewhere between the Turk and Greenblatt to Deep Blue or beyond. Even just a few months prior to Fan's defeat it was newsworthy if a go computer managed to beat a pro in a large handicap game and even us amateurs could handle most go programs.
The match will be starting soon, as I write this probably. Fan Hui is no slouch, true I think Lee would defeat him soundly, but the difference in strength between professional go players is not that wide a margin. Simply put if AlphaGo can beat one of them, I think it can probably beat them all. The go world is full of hubris, just like the chess world was, like Kasparov was before his match with Deep Blue. Even if AlphaGo wins, it will have at least one or two more matches before the majority of the go community is convinced. Initially Lee was very confident in his ability to defeat his silicon opponent, although Lee is showing some signs of humility now, unfortunately probably for the wrong reason. It sounds as if he has learned that the computer is capable of thinking like a human, and it is true that to tackle the sheer massive complexity of go, getting computers to think and behave more like humans was necessary, but don't forget about the massive computing power behind it - this is the computer's true strength. So in my view, the fact that the google team that made AlphaGo has given it a humanesque talent is really more of an oddity. Lee will be playing an alien intelligence, and he needs to prepare accordingly.
The funny thing is, go players have never really even developed anti-computer strategy - there was never a need. And after they become dominant I think there will continue to not be a need, computers were no match for us before and after this we simply won’t be a match for them.
If there were anti go computer strategy I have some predictions about what it might have looked like, and what I think Lee should try to do. See in go, the game simplifies in a way chess doesn't at least not to the same degree. Lee's greatest strength is going to be in the opening, he should avoid common opening themes that the computer can copycat off of and try to build a big lead, as the game winds down to the end the computer will become exceedingly strong. He probably should try to build strength and avoid complex tactical fights that depend on ko or furikawari. A point to his advantage, Fan Hui lost one game by only 2.5 points, a very close game and certainly not a crush, so Lee does have some chances. I saw an article about Fan Hui post-match, he has been sworn to secrecy! Although he did say that he thought Lee has a tough fight ahead of him and that he won’t be playing the same computer that played him – it’s gotten stronger!