Even if skills crucial to real-world domains are hard to train for, radical world transformation is achievable through verifiable R&D alone—spanning chips, fabs, robots, and further AI development—without needing AIs to master politics or negotiating.
Responding to Dwarkesh's skepticism about creating AIs that can master non-containerizable tasks, Ryan argues that if AIs can be extremely good at verifiable R&D—hardware, fabs, robots, and AI itself—that is sufficient for an industrial explosion that transforms the world without needing to be good at playing politics. ✦ AI generated
Ryan Greenblatt · Dwarkesh Podcast · 2026-08-11 · original ↗
plays this moment only · 39:47 — 45:00
“So let's step back and package this whole story... I think it's very plausible that it's very hard for GPT-8 to figure out how to make this transfer to those environments.”
If the AIs were really, really good at chip R&D, building fabs, orchestrating factories, designing robots, operating robots, and also at AI R&D — developing AIs for new downstream domains with whatever data is available — I think that would already be a pretty crazy situation. From there, you can get what we might call an industrial explosion, where the AIs are building out way, way more compute. Also, maybe you're already in a regime where AIs are doing huge amounts of R&D that humans have a hard time understanding.
verbatim transcript · starts at 39:47
39:47– Skills AI can’t train on: does it even need them?
39:47– Skills AI can’t train on: does it even need them?