Training set: 1. Image frame by frame from a AAA game 2. Human input frame by frame
AI reads the training set to fins the pattern between image and human input then it outputs "new images" for any new human input.
This would generate a game that clearly violates the Intellectual Property of the likeness of characters etc (Which can be solved by training an AI to modify those characters)
But the logic behind is fully "rewritten" and i believe would be original enough as to not violate any IP/Code's protection.
Why/why not?
Would this be like printing money?
[link] [comments]