Will we get "chess AGI" by 2030?

180Ṁ146

2030

80%

chance

ALL

I call "chess AGI" to an AI system that is superhuman in all chess-related tasks.

Computer are much better than human in playing chess. However, there are many examples for chess problem that humans can solve given some time, but computer can't handle. For example see the problems at https://chess.stackexchange.com/questions/23837/solvable-studies-that-computers-fail-to-crack

Resolution criteria:

Resolved to YES when a computer software is shown to be able to solve all of the chess problems (mate in n and start and win/draw) that an expert problem solver can reasonably solve within a day.

Resolved to NO at 1.1.2030 if no such software was shown to exist.

Technical AI Timelines

Chess

Board Games

Get

1,000

to start trading!

2 Comentários

7 Posições

10 Atividades

Ordenar por:

Does it have to give the correct evaluation or does it just need to play the optimal moves? (i.e. does it need to display 0.00 in fortresses or is it sufficient to be able to play moves that keep the draw?)

I am surprised to see the numbers that high with no NO betters. I didn't see any significant progress towards it in recent years, and there are some real gaps around proper reasoning vs. tree search and some nasty anti-computer chess problems out there