Now | David Quarel

It’s been a while!

Still in London!
Taught at ARENA-5 through ARENA-8. ARENA-9 is starting early October!
Now Head of Teaching at ARENA: lots more time to write new material
- Pretty happy with the new day on Monte Carlo Tree Search & AlphaZero I wrote.
Helping out ILIAD with development of their curriculum.
- Wrote stuff on RL, Solomonoff Induction and AIXI.
Spent the better part of a year at Timaeus, co-author on two papers.
Taught at AI Security Bootcamp in Vegas.

Started work at Timaeus as a Research Scientist: Does Singular Learning Theory mean anything in the Reinforcement Learning domain?

Still in London, still at LISA.
ARENA-4 just wrapped up a few weeks ago. Several students have written up the projects they worked on, including replications of papers and contributions to open-source projects.
Moved to a new flat in East London.

I live in London now, and I spend my time at LISA on various projects.
- Exploring causal interventions to explore the claim that Llama-2 “thinks” in English
- Use mechanistic interpretability techniques to work out the mechanism by which a LSTM trained to play sudoku does forward planning.
I’m the head TA at ARENA, a technical AI safety upskilling program.
Going to try again (perhaps in futility) to make a habit of writing more often here. Have a backlog of posts to clean and publish.
On ~May 2024 my book released!

I live in the UK now, and have been here since February
Currently attending/tutoring at ARENA
Collaborating with others at Cambridge AI Safety Hub
Almost finished a book I’m co-authoring, to be published soon(?)
Finally getting around to making a website now that large language models are good enough to hold my hand through the process