Hacker News Reader

These interviews are very, very good. Hearing these people talk casually in their own idiom about their work is the most efficient way I know of to get a feel for how things work at the cutting edge inside the big labs. That is useful whether you like what they’re doing or want to oppose it effectively.

Interview with the same two people from a year ago: https://www.dwarkesh.com/p/sholto-douglas-trenton-bricken

View on HN

They’re not actually talking casually, as OpenAI and Anthropic run a tight ship around PR. If either of these two say something the PR team doesn’t like it gets edited out.

Trenton and Sholto are very much “talking their book”. They’re doing it well, but it’s highly filtered and partial chat.

source: know a podcast episode which got removed because an OpenAI employee used “black box” to refer to NNs.

That’s interesting, though it doesn’t undermine the point that it’s information-dense, especially if you are coming at it from outside the field.

Reply on news.ycombinator.com