scripod.com

Dario Amodei — "We are near the end of the exponential"

Dwarkesh Podcast

Shownote

Dario Amodei thinks we are just a few years away from AGI — or as he puts it, from having “a country of geniuses in a data center”. In this episode, we discuss what to make of the scaling hypothesis in the current RL regime, why task-specific RL might lead...

Highlights

Dario Amodei, CEO of Anthropic, shares his grounded yet ambitious perspective on the rapid advancement toward artificial general intelligence—framing it not as a distant singularity but as an imminent, economically transformative force already reshaping industries and institutions.
02:28
RL scaling now shows similar patterns to pre-training scaling across diverse tasks, including math contests.
25:26
If we had an AGI-level 'country of geniuses in a data center', it would be obvious.
43:46
AI video editing capability expected in 1–3 years
46:20
Over-investing in data centers too early can be ruinous
1:19:32
Robotics revenue will reach trillions but progress won't be infinitely fast
1:37:23
Federal preemption—not a moratorium—is needed for effective AI regulation
2:08:23
Anthropic’s AI constitution is developed through three levels of iteration: within the company, across AI firms, and with society via public input projects.

Chapters

What exactly are we scaling?
00:00
Is diffusion cope?
12:36
Is continual learning necessary?
29:42
If AGI is imminent, why not buy more compute?
46:20
How will AI labs actually make profit?
58:49
Will regulations destroy the boons of AGI?
1:31:19
Why can’t China and America both have a country of geniuses in a datacenter?
1:47:41

Transcript

Dwarkesh Patel: So we talked three years ago. I'm curious, in your view, what has been the biggest update of the last three years? What has been the biggest difference between what it felt like last three years versus now? Dario Amodei: Yeah, I would say,...