Why AI researchers are questioning whether reward maximization is the wrong objective — Markdown | type0 | type0