✳flâneur — a map of the web's best reading

Questions about the Future of AI - by Dwarkesh Patel

dwarkesh.com · 6,773 words · saved by 2 readers

What started as an attempt to consolidate some thoughts from the last few interviews on my podcast has turned into this 6,000 word clusterfuck of questions and considerations. If you’ve got answers, or ideas for more questions, I’d be keen to read them in the comments below. I may compile a ‘Highlights from comments’ blog post or podcast episode in the future. Why don't we have reliable agents yet? Is agency training just a veneer of some MCTS-like scaffolding on the knowledge & intuition that pre-training gives you? Or is it much more difficult to develop? Here’s a case for agency being difficult to develop: Morevac’s Paradox. Evolution has been optimizing us for hundreds of million of years for being able to act like a coherent goal seeking agent even in the face of super dynamic environments, whereas evolution has spent at most hundreds of thousands of years optimizing us for language skills and abstract reasoning. So it's not that surprising that we got expert-level AI mathematicia

Blog Questions about the Future of AI Considerations about economics, history, training, deployment, investment, and more Dwarkesh Patel Apr 21, 2025 356 54 48 Share What started as an attempt to consolidate some thoughts from the last few interviews on my podcast has turned into this 6,000 word clusterfuck of questions and considerations. If you’ve got answers, or ideas for more questions, I’d be keen to read them in the comments below. I may compile a ‘Highlights from comments’ blog post or podcast episode in the future. Contents Capabilities Agency RL Idiot savants New training techniques P

Explore this link on the map →

saved by

related reading