Post #3343371

The obsession with utility-maximizing Bayesian agents only gets weirder as time goes on. You couldn’t really get further from the paperclip maximizer concept than an LLM, and yet they still won’t shut up about unaligned goals and hidden utility functions.