Trenton Bricken

The models are under-parameterized.

We're asking them to do a very hard task.

And they want to learn.

The gradients want to flow.

And so they're learning more general skills.

It goes without saying luck, obviously.

And I feel like I've been very lucky in the timing of different progressions has been just really good in terms of advancing to the next level of growth.

I feel like for the interpretability team specifically, I joined when we were five people.

We've now grown quite a lot.

But there were so many ideas floating around and we just needed to really execute on them and have quick feedback loops and do careful experimentation that led to signs of life and have now allowed us to really scale.

And I feel like that's kind of been my biggest value add to the team.

Which it's not all engineering, but quite a lot of it has been.

Yeah, yeah.

And this is why it's not all engineering, because it's running different experiments and having a hunch for why it might not be working and then opening up the model or opening up the weights and what is it learning?

Okay, well, let me try and do this instead and that sort of thing.

But a lot of it has just been...

being able to do very careful, thorough, but quick investigation of different ideas or theories.

I don't know.

I feel like I work quite a lot, and then I feel like I just am quite agentic.

If your question's about career overall, and I've been very privileged to have a really nice safety net to be able to take lots of risks, but I'm just quite headstrong.