

Ex-OpenAI CEO Says AI Labs Are Making a HUGE Mistake — Emmett Shear
6 snips Sep 26, 2025
Emmett Shear, former CEO of OpenAI and co-founder of Twitch, now leads the startup Softmax. He passionately discusses the existential risks posed by AI, asserting that current approaches may be dangerously misguided. Emmett highlights the importance of forming relationships with AI rather than treating them merely as controllable tools. He also critiques the unfair dismissal of doom proponents and underscores the need for open debates on alignment. His endorsement of a new book aims to provoke serious public discourse about these critical issues.
AI Snips
Chapters
Books
Transcript
Episode notes
Baseline Risk Argument
- Emmett Shear calls If Anyone Builds It, Everyone Dies the baseline argument people must grapple with when thinking about AI safety.
- He warns that ramping capabilities tied to goals leads to potentially dangerous outcomes as systems get smarter.
Humanity-as-Foom Analogy
- Shear draws an analogy between AI foom and humanity's impact on other species to illustrate sudden capability jumps.
- He argues such rapid improvement can upend the world in unpredictable, dangerous ways.
Engage With Assumptions
- Shear attributes some disagreement with Eliezer to an engineer’s versus scientist’s mindset.
- He urges people to specify which assumptions they reject rather than dismiss the argument out of hand.