search

carykh / PrisonersDilemmaTournament

Watch This Place's awesome video about iterated Prisoner's Dilemma for context! https://www.youtube.com/watch?v=BOvAbjfJ0x0
MIT License
205 stars 159 forks source link

ML & AI should not be allowed!!! #48

Closed Vaibbs closed 3 years ago

Vaibbs commented 3 years ago

ML and AI should not be allowed! The thing is ML and AI are not strategies and as far I know (think) we are supposed to submit strategies that anyone (or anything) can adapt and change and update, But ML (and AI) are like creatures (in this case the network) that have mastered the thing (the game), which is not same... But anyways tell me if I am wrong!!!

HannesGitH commented 3 years ago

ML/AI is just fancy matrix calculations, there is no magic creature behind it. It's still deterministic and not cheating at all, i would guess you wouldnt even find a very good AI solution since you cannot train on your opponents stragtegies.

Nevertheless any libraries besides numpy and random are banned anyways. (you could implement your own ML tho)

Vaibbs commented 3 years ago

Hmm... OK, also thank you for clearance!

HannesGitH commented 3 years ago

if you gather all implementations from the different forks to train an AI the story will probably be different and give you an actual chance at winning

if i had the time i might have tried