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 already-clean.goodfire.com/

Wow. Finished reading Emergence this weekend -- the memoir of @SussilloDavid. I could do a long tweet thread about all the ways in which I loved reading this. But I'll just say that I recommend this book, like really really recommend.

It’s moving and remarkable; there are clear parallels with @tarawestover’s hit memoir, “Educated,” which similarly exposes the deep fault lines of social class and poverty in the US. And like Westover, Sussillo can sure write. You will have a hard time putting this one down.

Some branches of neuroscience (and science in general) are unlikely to see real progress for a long time. This stagnation stems from an academic echo chamber that very effectively silences dissent. young researchers introducing new ideas are often marginalized or driven out before they can establish themselves. There must be quite a few of these people who left academia this way and are now working in industry. The current institutional framework simply won't fund or validate anyone bold enough to challenge entrenched ideas. It's a conformist paradise gated by tenure. BTW these stagnant areas tend to have dominant senior people who have never been wrong and shall remain "authorities" forever.

Why does everybody want managers to be ICs? Please someone explain this to me from first principles.