AI ALIGNMENT FORUMTags
AF

Interpretability (ML & AI)

•

Applied to AXRP Episode 19 - Mechanistic Interpretability with Neel Nanda by DanielFilan at 9h

•

Applied to ChatGPT: Tantalizing afterthoughts in search of story trajectories [induction heads] by Bill Benzon at 1d

•

Applied to More findings on maximal data dimension by Marius Hobbhahn at 2d

•

Applied to More findings on Memorization and double descent by Marius Hobbhahn at 3d

•

Applied to No Really, Attention is ALL You Need - Attention can do feedforward networks by Robert_AIZI at 4d

•

Applied to Mechanistic Interpretability Quickstart Guide by Neel Nanda at 4d

•

Applied to Spooky action at a distance in the loss landscape by Jesse Hoogland at 8d

•

Applied to [RFC] Possible ways to expand on "Discovering Latent Knowledge in Language Models Without Supervision". by gekaklam at 10d

•

Applied to How-to Transformer Mechanistic Interpretability—in 50 lines of code or less! by Stefan Heimersheim at 11d

•

Applied to List of links: Formal Methods, Embedded Agency, 3d world models, and some tools by thegearstoascension at 12d

•

Applied to Deconfusing "Capabilities vs. Alignment" by RobertM at 12d

•

Applied to Large language models learn to represent the world by Andrea_Miotti at 13d

•

Applied to Transformer Mech Interp: Any visualizations? by Raymond Arnold at 17d

•

Applied to 200 COP in MI: Studying Learned Features in Language Models by Neel Nanda at 17d

•

Applied to Reflections on Trusting Trust & AI by Itay Yona at 19d

•

Applied to Neural networks generalize because of this one weird trick by Jesse Hoogland at 19d

•

Applied to How does GPT-3 spend its 175B parameters? by Raymond Arnold at 22d

•

Applied to Can we efficiently distinguish different mechanisms? by Multicore at 22d