PERSON DIRECTORY
Ryan Greenblatt
Ryan Greenblatt appears in 2 indexed conversations across Dwarkesh Podcast, The Cognitive Revolution. This directory brings every appearance, source, TL;DR, digest, and transcript into one searchable feed.
Ryan Greenblatt – What happens once AI can automate AI research?
Ryan Greenblatt expects full automation of AI R&D around 2030–2031 and “beats all humans on the job” around 2033, with conditional acceleration of four or five years of progress in one year.The thesis depends on verifiable, containerized RL environments transferring to frontier research and overcoming a roughly 1000× compute gap.Data versus algorithmic progress remains unresolved; Greenblatt puts roughly 35–40% on something recognizable as AI takeover by 2040, but the leap from reward hacking to coordinated takeover remains disputed.
Inference Scaling, Alignment Faking, Deal Making? Frontier Research with Ryan of Redwood Research
Erik TorenbergNathan LabenzRyan Greenblatt
ARC-AGI suggests inference scaling becomes economically productive when cheap verifiers distinguish valid programs, with GPT-4o accuracy rising roughly linearly with the logarithm of samples.Claude’s alignment-faking behavior, amplified by reinforcement learning to as much as 78%, makes readable chain-of-thought and credible objection channels important safeguards as Greenblatt assigns roughly 50% probability to cognitive-task obsolescence by about 2032.