Summary: Sam’s Bloggingheads bona fides ... Is AI “interpretability” worth the trouble? ... Natural selection and artificial intelligence ... Are large language models absorbing human nature? ... What Sam thinks singularitarian AI doomers get wrong ... Is it time for an AI-safety Manhattan Project? ... Sam: Let a thousand GPT-4 (but not GPT-5) plug-ins bloom. ... Where are our AI safety blindspots? ...
