Let's face it, I'm very bad at chinese checkers. And my wife is pretty good at it. So I got very frustrated and decided to write an AI to beat her just once. From a transhumanist point of view, the AI that I wrote is an extension to myself.
So it is fair to say that now I can also beat my wife at chinese checkers.
(23/07/20: The final match is still to be played, but she actually defeated a work-in-progress version of the AI once...) (21/01/23: There was a never a final match, she continued winning all subsequent games and I decided to act as if those games never happened)
Sorry, this is not as exciting as GPT-3. I did this while on parental leave. You can't get too fancy with a baby crying.
This is just a very simple engine that uses the typical min-max tree with alpha-beta pruning, plus a transposition table to speed up things. But the code was profiled and optimized quite a bit.
You will need Python >= 3.8
(if on linux - maybe a bunch of libraries as a requirement to pygame like apt-get install python-dev libfreetype6-dev libsdl-image1.2-dev libsdl-mixer1.2-dev libsdl-ttf2.0-dev libsdl1.2-dev libsmpeg-dev)
make create_env
make run
Mouse input through pygame is a bit rusty, so you might have to repeat sometimes the move that is intended.