A nice article in Vox today, introducing some of the foundational puzzles facing AI interpretability researchers, by celia ford.

Reminder to submit proposals for access to the #NDIF 405b pilot access program, by tomorrow! ndif.us/405b.html

As AI models become more powerful, auditing them for safety & biases is crucial — but also challenging & labor-intensive. Can we automate and scale this process? MIT CSAIL researchers introduce "MAIA," which iteratively designs experiments to explain AI systems' behavior:

Very excited to see the launch of applied AI interpretability startup @goodfireAI by Tom McGrath and Eric Ho and Dan Balsam. The ndif.us team is collaborating with their technical efforts, which take advantage of the nnsight.net API.

A huge thank you to everyone who made #NEMI2024 a success! 🎉It was an honor to be the student lead organizer for this dynamic event, with insightful discussions and great networking. Special thanks to David Bau, Max Tegmark, Jannik Brinkmann, Kenneth Li, and Eric J. Michaud! #AI

Thanks Tal! 📜 In this paper, we provide a theoretically grounded review of causal (which, imo, ⊇ mechanistic) interpretability. We argue that this gives a more cohesive narrative of the field, and makes it easier to see actionable open directions for future work! 🧵

🚀 New NNsight features launching today! If you’re conducting research on LLM internals, NNsight 0.3 is now available. This update introduces advanced features, offering deeper insights for complex investigations into model behavior. 👇 Here’s what’s new: colab.research.google.com/github/ndif-te…

What is the goal of interpretability in AI? I spoke a bit about this at the recent far.ai alignment workshop: youtube.com/watch?v=AIfmSx… The event had excellent talks from many others, worth a view, linked in the thread.

"As someone who develops against LLMs, interpretability and transparency are everything to me" OpenAI is strangling their own ecosystem. Any AI companies that are serious about building a platform need to understand that transparency is the oxygen that sustains innovation.