"The Cognitive Revolution" | AI Builders, Researchers, and Live Player Analysis It's Crunch Time: Ajeya Cotra on RSI & AI-Powered AI Safety Work, from the 80,000 Hours Podcast
137 snips
Apr 11, 2026 
Ajeya Cotra, an AI researcher and forecaster known for work on AI timelines and safety strategy, explores the “crunch time” when AI could speed up its own progress. She digs into recursive self-improvement, automation loops, transparency and early warning signals, redirecting AI labor toward defense, and the bottlenecks that could make or break a safer path.
AI Snips
Chapters
Transcript
Episode notes
AGI Means Wildly Different Things To Different People
- Ajeya Cotra argues many AI disagreements hide a definition problem because people say “AGI” but mean anything from GPT-5-level usefulness to full human replacement.
- She contrasts panelists predicting AGI by 2030 with the same panelists expecting net job creation over the next decade.
The Core Fight Is About How Fast The World Changes
- Ajeya Cotra says the real divide is whether AI merely sustains normal progress or compresses something like 10,000 years of change into decades.
- Her far end is a 2050 world as alien to us as today is to hunter-gatherers, driven by automating intellectual work.
The Bigger Feedback Loop Includes Chips And Robots
- Ajeya Cotra expects “top human expert dominating AI” to start a broader automation loop, not just automate coding or research tasks.
- She thinks such systems could direct humans, then robots, to build chip fabs and physical actuators, closing the full stack of AI production.