Your network blocks the Lichess assets!

lichess.org
Donate

Science of Chess: The AI Revolution Hasn't Made Chess Better (yet!)

About the universal path to mastery - if there is a universal algorithm (like used in alpha zero) that can be demonstrated, but yet the mind can not replicate the results of the machine, what are the implications of how the mind works?

I think this is a wonderful question. My very-partial answer would be that this discrepancy highlights some of the heuristics the mind relies on that have great utility for everyday contexts, but can be hindrances for a formal setting like chess. It's why I find the difference between "natural" moves and "computer moves" so fascinating: I'd love to know more about what makes a natural move seem natural from a perceptual, cognitive, and psychological standpoint.

> > About the universal path to mastery - if there is a universal algorithm (like used in alpha zero) that can be demonstrated, but yet the mind can not replicate the results of the machine, what are the implications of how the mind works? I think this is a wonderful question. My very-partial answer would be that this discrepancy highlights some of the heuristics the mind relies on that have great utility for everyday contexts, but can be hindrances for a formal setting like chess. It's why I find the difference between "natural" moves and "computer moves" so fascinating: I'd love to know more about what makes a natural move seem natural from a perceptual, cognitive, and psychological standpoint.

Really interesting article. Thanks a lot for putting in the time to research and write this.

Really interesting article. Thanks a lot for putting in the time to research and write this.

Thanks for the interesting article!
As someone with a bit of knowledge of Go, the fact, that in Go there was a jump when the first Go-engine defeated the best humans, while in chess there was no noticeable spike, might have a reason.
When AlphaGo came around, it showed players, that some of their conceptions of good opening sequences (Joseki) were slightly off. Usually that was not by a lot, the masters of the past got them down quite well. But some turned out to be slightly better or slightly worse than humans believed for decades or centuries (e.g. the early 3-3 invasion versus star point openings). And even if they are maybe just 1 point or so off, being 1 point off a handful of times during the game can lead to an advantage, that professional players can convert to an advantage.
In chess however, the changes to high-end opening preparation were less drastic. Some variations might have fallen out of favour, but rarely did highly regarded openings completely disappear or completely new ones arrive out of nowhere. A Najdorf Sicilian or a closed Ruy Lopez were solid opening before and they still are after. The specifics in deep preparation might have changed, but at some point even super GMs come out of book and have to play the game instead of relying on computer assisted preparation.

Thanks for the interesting article! As someone with a bit of knowledge of Go, the fact, that in Go there was a jump when the first Go-engine defeated the best humans, while in chess there was no noticeable spike, might have a reason. When AlphaGo came around, it showed players, that some of their conceptions of good opening sequences (Joseki) were slightly off. Usually that was not by a lot, the masters of the past got them down quite well. But some turned out to be slightly better or slightly worse than humans believed for decades or centuries (e.g. the early 3-3 invasion versus star point openings). And even if they are maybe just 1 point or so off, being 1 point off a handful of times during the game can lead to an advantage, that professional players can convert to an advantage. In chess however, the changes to high-end opening preparation were less drastic. Some variations might have fallen out of favour, but rarely did highly regarded openings completely disappear or completely new ones arrive out of nowhere. A Najdorf Sicilian or a closed Ruy Lopez were solid opening before and they still are after. The specifics in deep preparation might have changed, but at some point even super GMs come out of book and have to play the game instead of relying on computer assisted preparation.

What an interesting and well-crafted article. Thanks for your efforts.

But let us not forget one thing: the Hippopotamus Opening is a lovely fresh spring for chess! The hedgehog moved us in the right direction and now the Hippopotamus (long may it reign) guides us toward the future, unerringly, like a a lover's cigarette glow in a dark room, or like the fine aroma of Texas barbecue wafting across a sunlit, summer field.

Let me be clear, however, that I am not promoting the use of tobacco. To the contrary, when it comes to smoking, I "just say no." But I would never try to control the behavior of one I loved, even for a moment.

And as I love the Hippopotamus, I would never try to rush it nor expand beyond it when it is not yet ready to expand.

What an interesting and well-crafted article. Thanks for your efforts. But let us not forget one thing: the Hippopotamus Opening is a lovely fresh spring for chess! The hedgehog moved us in the right direction and now the Hippopotamus (long may it reign) guides us toward the future, unerringly, like a a lover's cigarette glow in a dark room, or like the fine aroma of Texas barbecue wafting across a sunlit, summer field. Let me be clear, however, that I am not promoting the use of tobacco. To the contrary, when it comes to smoking, I "just say no." But I would never try to control the behavior of one I loved, even for a moment. And as I love the Hippopotamus, I would never try to rush it nor expand beyond it when it is not yet ready to expand.

I use the AI to explain the data collected by my chess cockpit from the inside of stockfish (classical and dynamic evaluation with SF 15.1). Very interesting is the interpretation of NNUE difference values.
https://www.amazon.de/Chess-Thinking-Age-Engines-Understanding-ebook/dp/B0GSH54B9F

I use the AI to explain the data collected by my chess cockpit from the inside of stockfish (classical and dynamic evaluation with SF 15.1). Very interesting is the interpretation of NNUE difference values. https://www.amazon.de/Chess-Thinking-Age-Engines-Understanding-ebook/dp/B0GSH54B9F

What has been brought to us by Stockfish and other engines? I think my first idea is the advance of the pawns a2 and h2 to a4 and h4. These are typical engine moves.

What has been brought to us by Stockfish and other engines? I think my first idea is the advance of the pawns a2 and h2 to a4 and h4. These are typical engine moves.