AI in Chess

How strong is Artificial Intelligence in chess now?
21 January 2021

Interview with 

Natasha Regan, Woman International Master


White chess set with black chess piece


Chris Berrow talks to Natasha Regan, Woman International Master (WIM) about the latest developments in chess AI.

Natasha - So it's very interesting. So this was built by DeepMind, and it was called AlphaZero because it had zero human knowledge in it. So the traditional engines might've been told the value of a Queen is such and such but AlphaZero was different. It was a sort of self-taught. So it didn't know the rules of chess. What it did was then play, 44 million games at lightning fast speed against itself starting completely random. So at first, most of the games would end in a draw because it didn't really know anything about chess and couldn't really make any progress, but by chance, some games would be decisive. And then what it would do is then look at the kind of moves that had played, where it had won and try and play a little bit more like that, and look at the moves where it's a lost cause it was playing against itself and try and play a little bit less than that. And so then it would kind of play more and more and then build up a picture of what patterns were good for it. And it was phenomenally successful.

Chtis - Does it spoil any of the games to have a computer that's almost unbeatable because I saw a quote from one, I think it's a Danish chess Grandmaster who said it's like playing against, uh, an alien, you know, it's just so good that you almost can't beat it as a human, but does that, does that ruin chess or does that become the aspiration for human players?

Natahsa - It's kind of, there's this, there's lots of things to that question. And this is kind of brought new possibilities that we didn't know about on the chess board. I mean, actually in chess for a little while, computers have been stronger than humans, so there was some famous much of Kasparov against deep blue. It was two year, two series match and Kasperov on the first and then Deep Blue on the second. So, and then after that computers have got a lot stronger. Um, so, so it kind of people are used to using computers as a tool to help them play tests as opposed to trying to, to beat them anymore, just because they're so strong. Actually in our online world now it's quite nice having the computers in terms of spectating chess, because you've got a lot of chess going on online at the moment and what you can do, watching a big tournament is if you want to see whether someone's winning the game, you can have a look kind of at the computer evaluation and see how strong the position is. 



Add a comment