This is an extremely lengthy post by Nick Cammarata.

Author's summary on twitter:

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

Roon's quote tweet

when nick told me that he had some interp breakthroughs, he was uh, underselling it. i think this may be remembered as a major moment in scientific history. news so good even eliezer will have to lower his pdoom. we can start making alignment into a 'normal science'

I feel like my understanding of what a neural net is has been catapulted into the future by five years or something. it has brought back a lot of the sheer joy I had for studying the mystery of deep learning. neurons are already clean! they form a wonderful alien ontology!

I'd love to know:

How much of a real advance is this, vs things we already knew?
How is this likely to be useful in practice for the purpose of alignment?

fluke32

fluke32