Joseph Bloom

Joseph Bloom is a mechanistic interpretability researcher who co-founded Decode Research / Neuronpedia and led development of the SAELens library for training sparse autoencoders. He is now Head of White Box Evaluations (model transparency) at the UK AI Security Institute, and has co-authored prominent work including Open Problems in Mechanistic Interpretability, SAEBench, and "Sparse Autoencoders Do Not Find Canonical Units of Analysis." Before joining AISI he worked as an independent AI safety researcher focusing on decision transformer interpretability.

No endorsements yet.

Personal Website

https://www.jbloomaus.com/

LinkedIn

LinkedIn

Twitter / X

-

LessWrong

-

EA Forum

-

No grants recorded.