Why modelling multi-objective homeostasis is essential for AI alignment (and how it helps with AI safety as well). Subtleties and open challenges.
by Roland Pihlakas and Three Laws
I notice that there has been very little discussion on why and how considering homeostasis is significant, even essential for AI alignment and safety. The current post aims to contribute to or assamending that situation. In this post I will treat alignment and safety as explicitly separate subjects, which both...
Jan 12, 202548