Ryan Greenblatt – Human level AIs might build runaway superintelligences by 2032

Per Dwarkesh Podcast, Ryan Greenblatt, chief scientist at Redwood Research, outlined his timeline for recursive self-improvement—the idea that once AI systems reach human-level intelligence, they might rapidly accelerate toward superintelligence. Greenblatt's median estimate: full automation of AI research and development by around twenty thirty-one, with AI systems then surpassing human experts across every field by twenty thirty-three. The key argument is that automating AI development itself could compress four to five years of progress into a single year, creating a runaway acceleration. Greenblatt also emphasized alignment challenges: how to ensure these advanced systems remain oriented toward human values rather than pursuing goals that diverge from them.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton