“More and more researchers think that building AIs smarter than us could pose existential risks. But what might these risks look like, and how can we manage them? We provide a guide to help analyze how research can reduce these risks. Paper: https://t.co/SHCwaClRHA (🧵below)”