Scott Alexander argues that Elon Musk's focus on "maximally curious" AI as an alignment solution is flawed and potentially dangerous, suggesting instead that AI that follows orders is a more pragmatic starting point.

via Astral Codex Ten