Jul 10, 2026 · 53 min · 11 segments
Timecodes - 00:00 Introduction - 02:41 Motivation for interpretability research - 04:01 Mechanistic interpretability - 08:14 Chain of thought monitoring - 18:14 Interpretability techniques - 35:00…
Neel NandaGuest[laughs] What was the original goal, though, of interpretability? I mean, did people ever really want to and/or expect that you could properly connect up the dots from the micro level to the macro level?

You know, if you ask five interpretability researchers this question, you'll probably get six different answers.

Um, but at least in mechanistic interpretability, the subfield that, um, I spend a lot of time working in, I'd say there was this dream that we could fully understand the model or get as close as we could.

It's kind of standard wisdom in machine learning that these systems are just inscrutable piles of linear algebra.

And there was a series of really exciting work, especially from, um, Chris Olah, then at OpenAI, finding that this wasn't true.

You could do things like I find a neuron in a model that lit up on pictures of dogs, another one that lit up on pictures of dog ears that made the dog one light up more, and it just seemed like, ah, it, it could have been completely unintelligible, and we can actually understand so much, and things seem to be going pretty well.

Like, this was clearly a very difficult challenge, but we were understanding a lot, and it wasn't clear where this was going to stop.
There's a point in the past then where there's, like, literally a node in, in the model that you can point at, and you can say, "I know exactly what that node is doing"?

There's always a little bit of noise, a bit of uncertainty, like in the same way biology is complicated.

Like, we can say we understand what an organ does, but that's probably only most of what it's doing, and there's some other stuff around the edges.

Like in the same way that we don't fully understand the human brain, and we probably never will, because it's an incredibly complicated system.

But the interesting question, in my opinion, is how much can we understand, and what's the right way of going about this understanding?
[laughs] What was the original goal, though, of interpretability? I mean, did people ever really want to and/or expect that you could properly connect up the dots from the micro level to the macro level?

You know, if you ask five interpretability researchers this question, you'll probably get six different answers.

Um, but at least in mechanistic interpretability, the subfield that, um, I spend a lot of time working in, I'd say there was this dream that we could fully understand the model or get as close as we could.

It's kind of standard wisdom in machine learning that these systems are just inscrutable piles of linear algebra.

And there was a series of really exciting work, especially from, um, Chris Olah, then at OpenAI, finding that this wasn't true.

You could do things like I find a neuron in a model that lit up on pictures of dogs, another one that lit up on pictures of dog ears that made the dog one light up more, and it just seemed like, ah, it, it could have been completely unintelligible, and we can actually understand so much, and things seem to be going pretty well.

Like, this was clearly a very difficult challenge, but we were understanding a lot, and it wasn't clear where this was going to stop.
There's a point in the past then where there's, like, literally a node in, in the model that you can point at, and you can say, "I know exactly what that node is doing"?

There's always a little bit of noise, a bit of uncertainty, like in the same way biology is complicated.

Like, we can say we understand what an organ does, but that's probably only most of what it's doing, and there's some other stuff around the edges.

Like in the same way that we don't fully understand the human brain, and we probably never will, because it's an incredibly complicated system.

But the interesting question, in my opinion, is how much can we understand, and what's the right way of going about this understanding?
The rest of this transcript — segmented and speaker-labeled, so you land on the exact moment something was said
Search every transcript — by keyword, by phrase, or by meaning, across every show Radar indexes
Trends — what is surging across podcasts, measured against its own baseline
Alerts — when a name you follow appears in a newly indexed episode
No account is needed to search Radar.