Hacker News new | ask | show | jobs
by vegarab 2413 days ago
This is my point exactly. The model is trained without any prior domain knowledge at all. It only has access to a game world where the constrains in the world is a representation of the game's rules.