It’s the biggest chess news of the year. Perhaps the biggest chess news of the century. You might even consider it the biggest chess news of all time. Nigel has already written about this, but I think it’s worth another article.
The games we’ve seen so far have been fascinating and totally unlike human games. The choice of openings is the first point of interest. AlphaZero seems to prefer queen’s pawn or flank openings (1. d4, Nf3 or c4), disagreeing with Fischer’s dictum that 1. e4 is ‘best by test’. It doesn’t seem to think much of Black’s sharper defences such as the Sicilian and the King’s Indian. It liked the French for a time before switching to the Caro-Kann and then 1… e5, choosing the Berlin Defence against the Ruy Lopez. At the same time, several games featured positional sacrifices, demonstrating a preference for initiative over material. No doubt it had worked everything, or at least almost everything, out: it wasn’t just being speculative.
So already, after teaching itself in only four hours, it must be pretty close to playing perfect chess. How well will it play after 4000 hours?
It was also interesting, or perhaps disturbing, to read here that, of the sixty games so far completed in the 1st English Correspondence Chess Championship, fifty eight have been drawn. These days, because engine assistance is permitted, the vast majority of correspondence games result in the point being shared. The combination of human brain and computer brawn is starting to approach perfection, but still a long way short of AlphaZero. Compared with this, the number of decisive games in the London Chess Classic (10 out of 45 after a late flurry of excitement in rounds 7 and 9) seems positively thrilling.
What impact will this have on chess between humans? At amateur level, playing blunder-strewn games in inter-club matches and weekend congresses, very little. If AlphaZero becomes available online in some form I guess it will, sadly, mean the demise of correspondence chess. It will also have a big impact on top level chess, quite probably leading to more draws than today. Professional players will be able to carry out deeper research further into the game. People have been predicting the death of chess for more than a century: perhaps AlphaZero demonstrates how the chess world will end. Not with a bang but a whimper.
There are answers, though. Some pundits are predicting the rise of Chess960, while others, and I’d probably put myself in their camp, believe that using different starting positions destroys the purity of chess. I don’t think I’d be opposed to the occasional Chess960 tournament, though. We’ll no doubt see more tournaments at faster time limits, which are also more entertaining for spectators. Perhaps we’ll see more invitations for creative players like Rapport and Jobava rather than the ‘bore draw’ specialists.
I’m currently reading Yuval Noah Harari’s books Sapiens and Homo Deus. Harari predicts that homo sapiens will, in the not too distant future, die out, to be replaced by immortal cyborgs. I suppose that, in one sense, AlphaZero is a step in this direction. I’m not entirely convinced by Harari’s arguments, or at least I hope I’m not, and I hope he’ll be proved wrong. Not that I’ll be around long enough to find out, though.
All this prompts thoughts about how we might change chess for the better, which I’ll come back to later, and how we might change society for the better, which is a topic for another time and place, although not unrelated to my views on chess, and, specifically, junior chess.
Meanwhile, here’s another video of one of the AlphaZero v Stockfish games for you to enjoy.