Nick Cammarata says modern agent tools make the vast majority of neurons in large models readable
AI summary
Researcher Nick Cammarata said that with modern agent tools and closer observation, the vast majority of neurons in models with parameter scales up to trillions are readable, circuit structures are clear, and many famous "entangled neuron" cases are actually misunderstandings. He was involved in OpenAI's mechanistic interpretability work in his early years and initiated OpenAI's first large-scale automated interpretability project; at the time, using raw tools, he found only about 2% of neurons readable, which made him pessimistic. He now believes that most of the obstacles he expected do not exist, and ambitious interpretability may be solved soon because automated scientists continue to improve.
interpretability might be the most important scientific problem in the world right now, as I don't think we can solve alignment if we don't know what we're building. I think I have good news about it. I co-authored early work in mechanistic interpretability at OpenAI and came away pessimistic, because neural networks appeared to fold into arbitrary tangled knots. years later, I started openai's first large-scale automated interpretability project, finding about 2% of neurons legible with the primitive tools we had
with modern agentic tools and a much closer look, the picture looks totally different. the vast majority in models up to trillions of parameters are legible, and their circuits beautiful. many famous examples of tangled neurons turned out to be misunderstood. The roadblocks I expected mostly aren't there. I now think ambitious interpretability will be solved by the time we need it most, and possibly quite soon, as automated scientists keep improving. it may be a bitter lesson for interpretability: the simple thing, at absurd scale, works
https://already-clean.goodfire.com/View the quoted post on X