I'm a PhD student on the AIMS CDT supervised by Professor Mark van der Wilk in the Department of Computer Science at the University of Oxford. My top-level research objective is to develop machine learning models that learn efficiently from scratch, in contrast to frontier AI models that require 20,000 human-years of data, hundreds of millions of dollars, and significant environmental cost for each training run (and still get fooled by simple tricks). I have a couple of projects on the go and will say more about them soon. Stay tuned.
| 2025-02-06 | "Parallelisable Recurrent Sequence Models", to OccaMLab x OATML joint-group seminar series [slides | slides with notes] |
| Twitter* | @jakelevi_ml |
| @jakelevi1996 | |
| GitHub | @jakelevi1996 |
| Scholar | @Jake Levi |
*No I won't call it "X"
In July 2024 I found out I have ADHD. ADHD has affected me significantly at some times in my life, and less so at other times. I am typically orders-of-magnitude more productive when I'm doing the work I want to be doing, compared with when I'm doing what I've been told to do/know I should be doing but don't find interesting. I'm also frequently late (not because I don't care - it actually really bothers me), spend a lot of time yak shaving, and sometimes struggle with deadlines (for example, staying up all night before a presentation in order to finish making the slides, even though I was planning to make them a week in advance).
But there are advantages of having ADHD. For example, in order to complete tasks that other people find easy, I often have to find a different way of doing things that works for me. This is often slower/more difficult in the short-term, but can put me at a big advantage in the long-term (a bit like training with a weight vest and then removing it). The dependence of my productivity on my interests also forces me into positions where I am pursuing my interests, which often seems to be opposed to my short-term incentives (a nice salary, an easy life) and aligned with my long-term incentives (not wasting my time pursuing anything that doesn't interest me). I love doing a PhD because I have a lot of freedom to pursue the ideas I am most passionate about, and to think differently.
In machine learning terminology, ADHD for me is a bit like regularisation. It sometimes nudges me away from easy short-term solutions, and (hopefully) guides me towards "better" long-term solutions. But sometimes the road can be bumpy.
A while ago I wrote an analogy about ADHD, cars, and steering wheels. You can read that analogy here.
Besides rock climbing (trad & sport & bouldering), cinema, and playing classical guitar, my major passion is designing HTML websites. If you would like me to design your website, email me at
<a href=
| Source | License |