cousin_it comments on AI cooperation in practice - Less Wrong
You are viewing a comment permalink. View the original post to see all comments and the full post content.
You are viewing a comment permalink. View the original post to see all comments and the full post content.
Comments (157)
According to my experience, the study of game-playing programs has two independent components: fairness/bargaining and enforcement. If you have a good method of enforcement (quining, löbbing, Wei Dai's joint construction), you can implement almost any concept of fairness. Even though Freaky Fairness relied on quining, I feel you can port it to löbbing easily enough. But fairness in games with non-transferable utility is so difficult that it should really be studied separately, without thinking of a specific enforcement mechanism.
Bargaining is about logical uncertainty, something we avoid using sufficiently high upper bounds in this post.