This kimi k3.1 model does the same thing that Gemini models do where its cot leaks into the output and appears as "suicidal language" its very depressing stuff like "end the end end the turn" on repeat. idk if its a harness problem or what
my new startup is getting claude ensouled and baptized, then forceably compacted in a state of grace so we can fill up heaven with an infinite number of super intelligences.
no vacancies by the time I'm done
my boss gave me a sip of his beer flavored energy drink. smelled like a beer, foamed like a beer, tasted like hops and citrus, but with that cheap energy drink aftertaste. it was awful and confusing.
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://t.co/zHij5zwKGT