hnowak

The following are some non-technical ideas about AI alignment based on human beliefs, rather than our true reward function. My impression is that the role of beliefs is often implied in passing, but I haven‘t found any elaborations on the topic. I‘d be grateful for relevant references, if someone knows...