Quote:
Originally Posted by Brokenstars
He was not talking about physical tells at all as when he starts talking about poker/solvers he specifically says, "and we can talk about chess the sameway". He doesn't understand the definition of gto.
https://youtu.be/bcGwHworAo0?t=3794
He says,
"The problem with GTO is that you can actually be very exploitative against somebody that's actually playing perfectly because the AI is perfected around what is the rational set of decisions in every spot and so you can set people up to make a lot of really bad mistakes and I think helmuth understands that and so because he is one of like this dying breed of people that plays live he's able to just be so exploitative"
I’m pretty sure that chess engines are an iterative process that changes its play from experience. The early bots were terrible, but as they play more and more they converge towards the correct strategy. Not saying this is the same for poker GTO.
I was watching a documentary on the Go AI, and it is very surprising how experimental it is and how it tests out different strategies seemingly just because it wants to try it out. Since AI’s are able to play tons of games before converging on the correct strategy, it doesn’t make sense to think about them as always making the most rational decisions. Some AI’s basically brute force what it thinks is the optimal strategy.
But I think that AI’s aren’t playing the perfect strategy in chess, or else they wouldn’t always be improving. They are just getting better and better.