Linkpost: Neural Networks Are Already Clean
Key takeaway
This is an extremely lengthy post by Nick Cammarata.
Why it's on PDOOM
PDOOM selected this story for its alignment & control signals: Alignment, Interpretability.
AlignmentInterpretabilityFrom LessWrong AI
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 au
By Yair Halberstadt