Dario Amodei on Dwarkesh Podcast
Dario Amodei — "We are near the end of the exponential"
A substantive, technically detailed interview, but much of its scaling, coding-timeline, diffusion, and governance material overlaps Amodei’s other appearances—and one corpus entry republishes this episode nearly verbatim. Worth hearing for the finer distinctions around learning and verifiability; casual listeners can skip.
But there is a puzzle either way, which is that in pre-training we use trillions of tokens. Humans don't see trillions of words. So there is an actual sample efficiency difference here. There is actually something different here. The models start from scratch and they need much more training. But we also see that once they're trained, if we give them a long context length of a million — the only thing blocking long context is inference — they're very good at learning and adapting within that context. So I don’t know the full answer to this.
Dario Amodei argues that AI capability progress remains a steep but smooth exponential, with a country of geniuses in a data center plausible within one to three years and highly likely within ten. He expects economic adoption, governance, and distribution of benefits to lag technical progress, making enterprise diffusion, safety regulation, political freedom, and global access central challenges.