Lawrence Chan

PhD Candidate

Berkeley AI Research

As of January 2023, I’m currently working at the Alignment Research Center doing evaluations of large language models. Previously, I was at Redwood Research, where I worked on adversarial training and neural network interpretability.

I’m also doing a PhD at UC Berkeley advised by Anca Dragan and Stuart Russell. Before that, I received a BAS in Computer Science and Logic and a BS in Economics from the University of Pennsylvania’s M&T Program, where I was fortunate to work with Philip Tetlock on using ML for forecasting.

My main research interests are mechanistic interpretability and scalable oversight. In the past, I’ve also done conceptual work on learning human values.

I also sometimes blog about AI alignment and other topics on LessWrong/the AI Alignment Forum.

Recent Publications

Bilal Chughtai, Lawrence Chan, Neel Nanda

February, 2023 arXiv preprint

A Toy Model of Universality: Reverse Engineering How Networks Learn Group Operations

We reverse engineer small transformers trained on group composition and use our understanding to explore the universality hypothesis.

Neel Nanda, Lawrence Chan, Tom Lieberum, Jess Smith, Jacob Steinhardt

January, 2023 ICLR 2023 (Spotlight)

Progress measures for grokking via mechanistic interpretability

We fully reverse engineer 32 small transformers trained on modular addition and use this understanding to study grokking.

Buck Shlegeris, Fabien Roger, Lawrence Chan, Euan McLean

December, 2022 arXiv preprint arXiv:2212.11281

Language models are better than humans at next-token prediction

We compare humans to small language models on next-token prediction tasks, and find that even relatively small language models consistently outperform humans.

Richard Ngo, Lawrence Chan, Sören Mindermann

December, 2022 arXiv preprint arXiv:2209.00626

The alignment problem from a deep learning perspective

We argue that AGIs trained in similar ways as today’s most capable models could learn to act deceptively to receive higher reward; learn internally-represented goals which generalize beyond their training distributions; and pursue those goals using power-seeking strategies.

Lawrence Chan, Adrià Garriga-Alonso, Nicholas Goldowsky-Dill, Ryan Greenblatt, Jenny Nitishinskaya, Ansh Radhakrishnan, Buck Shlegeris, Nate Thomas

December, 2022 AI Alignment Forum

Causal Scrubbing: a method for rigorously testing interpretability hypotheses

We introduce a more principled approach for evaluating the quality of mechanistic interpretations via behavior-preserving resampling ablations—converting hypotheses into classes of activations inside a neural network can be resampled without affecting behavior.

See all publications