Sunday, January 9, 2022

Chess Engines are Cheating Engines; Plato's Failed Academy

A good chess engine requires the engine to have played some millions of games and to evaluate 40,000+ positions per second. This isn't computers beating humans at chess, this is a computer flagrantly cheating at chess and still barely winning. 

Moreover, this is nothing more than calculating a lookup table on the fly. Faster to pre-calculate the "proven" best move on every board state and simply respond with that move every time. 

More precisely, the programmer of the chess engine could do everything the chess engine does with pencil and paper, playing a chess game that way, except the part where the programmer would take years to perform every turn. It was never man vs. machine, it was man vs. machine-mediated man. The only surprising thing about this is how powerful a machine the cheater needs before he wins. Not only does he need to perform millions more operations than his opponent, he often needs an entire team looking over his shoulder and kibitzing.

It is easy for a random asshole on the internet (me) to pwn chess programmers like this because they're not trying to play chess. They believe heaven is real, but instead of St. Pete it's Plato at the gates; they're auditioning to be accepted by the great Academy in the Sky. They don't want to play chess. They don't want to figure out how thinking works. They want to derive the Platonic "solution" of chess. They want to use all sorts of complicated math to show off how profound they are. 

I find this just a little revolting considering how many real problems they could be working on instead. E.g. refining the epistemic training regimen. E.g. fertilizing crops correctly. E.g. zero-pesticide planting that has pesticide-comparable yields. E.g. convincing cities not to look like fungal rot. E.g. if you insist on being profound and pointless, study consciousness. (Admittedly difficult since you merely find, laboriously, that Descartes was right.)

P.S. I will consider a chess engine to win chess when it's cost-effective to develop and build a chess engine to play chess instead of hiring some bloke. We have to imagine that chess is a wealth-creation endeavour instead of a hobby. Even if we get self-driving cars it's likely that, without subsidies, hiring some bloke to drive you will still be cheaper. 


No comments: