Scientists Analyzed 24,000 Chess Matches to Understand Cognition(nautil.us)
nautil.us
Scientists Analyzed 24,000 Chess Matches to Understand Cognition
http://nautil.us/blog/scientists-analyzed-24000-chess-matches-to-understand-cognition
9 comments
Chess is just a convoluted version of TicTacToe. The longstanding idea that it's somehow representative of human intellect is ridiculous.
I'd like to playfully add on to this. Chess is a convoluted version of tic tac toe the same way that a barbell is a heavier version of a handle, and to say opening a swing door is representative of human strength is ridiculous.
Maybe to follow up on that even, Chess is absolutely solvable the same way tic-tac-toe is solvable, and computers have been better than people for years, but it's an incredible playground for people to think against each other competitively and demonstrate whose ideas are better. Even on the computer chess side, there are so many ideas to explore from parallel programming to task scheduling to compression for creation of endgame table bases and reinforcement learning for modern evaluation functions used in open source projects like Leela Zero.
Is it solvable or has iterating on current approaches simply converged? There obviously has been breakthroughs in computers playing games recently
I thought I responded to this earlier, but it looks like I didn't, sorry! In short, it's solvable in theory. The game as it's currently played has a finite game tree (thanks in part to the 50-move rule) and that finite game tree can in theory be brute-forced. In practice, I think it's one of those problems that's so large we can't possibly solve it. For an idea of the scale of a complete solution, solutions for all positions with 2 kings and 4 other pieces occupy ~100GB and solutions for positions with 2 kings and 5 other pieces are ~10 TB and it's believed that 2 kings + 6 will be in petabyte scale.
As a pastime invented by humans and enjoyed by a big percentage of humans, why would it not be representative of human intellect?