[AINews] Agents disprove Polysemanticity; greatly improve Interpretability & Alignment

A version of Muse Spark Open was released today, fulfilling their original promise but with a model weaker than their Spark 1.3. It’s also worth watching the Tesla Roadster demo - not strictly AI but nothing cooler than real life flying cars. Our title story today is a rare one on interpretability and alignment. We last caught up with Goodfire in Feb: This is one of the most important results in AI, period. TLDR: neural networks are not tangled knots (polysemanticity is wrong). You can study their own neurons directly, without first "unfolding" them with a second model. 90% of neurons in Qwen-32B have one job. Nick Cammarata… 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 44 Replies · 71 Reposts · 1.5K Likes 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…