Anthropic CEO on Leaving OpenAI and Predictions for Future of AI

Key ideas

Amodei recounts leaving OpenAI with a group that shared views on scaling and safety, and explains how Anthropic weighed commercial timing, governance and public pressure.

From a machine transcript of the audio. Quotes link to the moment they are said.

1. 1

   Founding can start from accumulated opinions

   Amodei says he always saw himself as a scientist, not a founder or CEO. Watching decisions at several organizations, he says, showed him he had his own views on growing a company, running research and bringing its products out responsibly.

2. 2

   Leave with a group and a thesis

   Amodei describes seven co-founders and 14 or 15 people in total leaving OpenAI, which he puts at about 10% of its size at the time. He says they were a coherent group who believed in scaling language models and wanted to build them where safety principles held top to bottom, and he notes that OpenAI supported the scaling view.

3. 3

   Decide where principle overrides commercial timing

   Amodei says about 99% of Anthropic's decisions would match a normal company's, and that its public benefit structure matters in the remaining key ones. His example is holding back an early version of Claude in 2022 out of concern that a big public release might speed things up faster than the ecosystem could handle, a call he thinks was right overall but says is pretty debatable.

4. 4

   Have customers build for future models

   Amodei says he tells enterprise customers to start with what a model can do today while planning for where it will be in one or two years, since going from proof of concept to product-market fit takes about that long. He says that if a model does a task 40% of the time, it can probably reach 80 or 90% in that period.

5. 5

   Do not steer by crowd approval

   Amodei warns that chasing short-term approval, on social media or elsewhere, can make smart people act much dumber than they are, and can push them into one extreme or into playing both sides. He prefers to speak when he has something substantive to say, and expects his decisions to be judged more fairly five to ten years out.

In Dario’s words

What’s available here

• Original source

  Linked. Opens on YouTube.

• Player on this page

  Available. YouTube’s own player, loaded only when you press play.

• Full text or transcript

  Not reviewed. No license to show a text version here.

• Our summary

  Draft. Written by us with AI from a machine transcript of the video’s audio (Whisper) and checked against it. Machine transcripts can mishear; each quote links to the moment so you can hear it. Not yet rights-reviewed.

• Ask questions with AI

  Not reviewed. Off until there is licensed text to work from.