I am a DPhil student in Computer Science at the University of Oxford supervised by Yarin Gal.
I work on safe and secure machine learning through Bayesian deep learning.
Trustworthy systems ought to be uncertain, but much of modern machine learning is over-confident by design. My work aims to understand uncertainty and confidence in neural networks.
Recent publications/preprints, for a full overview see my Google Scholar page.
@article{farquhar_statistical_2020, title={On Statistical Bias In Active Learning: How and When to Fix It}, author={Farquhar, Sebastian and Gal, Yarin and Rainforth, Tom}, journal={International Conference on Learning Representations}, year={2021} }
@article{farquhar_liberty_2020, title={Liberty or Depth: Deep Bayesian Neural Nets Do Not Need Complex Weight Posterior Approximations}, author={Farquhar, Sebastian and Smith, Lewis and Gal, Yarin}, journal={Advances in Neural Information Processing Systems}, year={2020} }
@article{farquhar_radial_2020, title={{Radial} {Bayesian} Neural Networks: Beyond Discrete Support In Large-Scale Bayesian Deep Learning}, author={Farquhar, Sebastian and Osborne, Michael and Gal, Yarin}, journal={Proceedings of the 23rd International Conference on Artificial Intelligence and Statistics}, year={2020} }
@article{farquhar_single_2020, title={Single Shot Structure Pruning Before Training}, author={van Amersfoort, Joost and Alizadeh, Milad and Farquhar, Sebastian and Lane, Nicholas and Gal, Yarin}, journal={preprint}, year={2020} }
@article{farquhar_unifying_2018, title = {A {Unifying} {Bayesian} {View} of {Continual} {Learning}}, language = {en}, journal = {Bayesian Deep Learning Workshop at NeurIPS}, author = {Farquhar, Sebastian and Gal, Yarin}, year = {2018} }
@article{farquhar_towards_2018, title = {Towards {Robust} {Evaluations} of {Continual} {Learning}}, journal = {Lifelong Learning: A Reinforcement Learning Approach Workshop at ICML}, author = {Farquhar, Sebastian and Gal, Yarin}, year = {2018} }
@article{farquhar_differentially_2018, title = {Differentially {Private} {Continual} {Learning}}, language = {en}, journal = {Privacy in Machine Learning and AI workshop at ICML}, author = {Farquhar, Sebastian and Gal, Yarin}, year = {2018} }
@article{brundage2018malicious, title={The malicious use of artificial intelligence: Forecasting, prevention, and mitigation}, author={Brundage, Miles and Avin, Shahar and Clark, Jack and Toner, Helen and Eckersley, Peter and Garfinkel, Ben and Dafoe, Allan and Scharre, Paul and Zeitzoff, Thomas and Filar, Bobby and others}, journal={arXiv preprint arXiv:1802.07228}, year={2018} }
@article{farquhar_pricing_2017, title={Pricing Externalities to Balance Public Risks and Benefits of Research}, author={Farquhar, Sebastian and Cotton-Barratt, Owen and Snyder-Beattie, Andrew}, journal={Health Security}, volume={15}, number={4}, pages={401--408}, year={2017}, publisher={Mary Ann Liebert, Inc. 140 Huguenot Street, 3rd Floor New Rochelle, NY 10801 USA} }
Before my DPhil, I led the Global Priorities Project – a joint project of the Centre for Effective Altruism and the Future of Humanity Institute at the University of Oxford – which connects research from the University with the realities that policy-makers face in order to offer evidence-based non-partisan policy proposals. GPP focused on technological risk management and frameworks that encourage evidence-based policy-making.
I was Acting Executive Director of the Centre for Effective Altruism in 2015 – which works in Oxford to use evidence and analysis to make the world a better place, and runs Giving What We Can, 80,000 Hours and the Global Priorities Project in addition to building a community of effective altruists.
At McKinsey, I developed strategy for leading European public sector providers of services in health and social care, and helped to implement large-scale change programmes in local areas serving more than 2 million people.
Before McKinsey, I was on the team that set up 80,000 Hours – a social enterprise which helps people have a meaningful ethical career which now reaches over 1 million people a year, raised over $5 million in funding, and been featured by the Washington Post, the BBC, TED, and NPR among others.
Site based on layout by Joost van Amersfoort.