Colaboration with Vedran Dunjko (Leiden) and Scott Aaronson (UT Austin) | 2022 | In this paper we provide the strongest complexity theory evidence so far that quantum computers can compute functions that classical computers cannot. We also provide the strongest evidence that a number of quantum supremacy experiments are classically hard. The paper will be released in mid-November.
Collaboration with Jan Kirchner (OpenAI) | 2023 | We applied information theory and neuroscience techniques to analyse how large language models encode information and protect it from noise. Providing insight into the interpretability of models.
Alignment Research Center/METR with Beth Barnes under the MATS program | 2022 | We developed and ran evaluations of large language models. This project helped set the groundwork for the current interest in model evaluations.
Universiteit Leiden | 2020-2024 | Mainly focused on topics in complexity theory, circuit cutting and quantum machine learning.
University of Warwick | 2016-2020 | Thesis in computational fluid dynamics, graded 'high/exceptional first', 1st class in every year.
Programming Almost all my PhD code has been in python, normally with TensorFlow but less frequently with PyTorch. I used a C++-derived language (Basilisk) for my Masters research and C for high-performance computing.
Human English - native, Dutch - A2.
University Officer Training Corps | UOTC Birmingham | The UOTC is provides reserve officer training fitted around university for university students.
AI safety fundamentals facilitation | I facilitated one round the AI safety fundamentals course in 2022.
Academic Email: s.c.marshall@liacs.leidenuniv.nl.
Generic Email: s.c.marshall98@gmail.com
I do not list references on my website for obvious reasons, feel free to reach out and request some.