Sotabase

Career

· Research Scientist, DeepMind2022–
· Junior Research Fellow, Christ Church, University of Oxford2021–2022
· Research Scientist Intern, DeepMind2021–2021
· Project Manager, Future of Humanity Institute2017–2017
· DPhil, University of Oxford2017–2022
· Director, Global Priorities Project2015–2017
· Acting Executive Director, The Centre for Effective Altruism2015–2016
· Business Analyst, McKinsey & Company2013–2015
· Director Of Message And Market Research, Centre for Effective Altruism2011–2013
· Summer Business Analyst, McKinsey & Company2011–2011
· Summer Intern, Record Currency Management2010–2010
· Summer Intern, Metaswitch Networks2009–2009
· MPhysPhil, Physics and Philosophy, University of Oxford2008–2012

Publications (42)

Detecting hallucinations in large language models using semantic entropy
Nature · 2024
879
cited
The Malicious Use of Artificial Intelligence: Forecasting, Prevention, and Mitigation
arXiv.org · 2018
837
cited
Semantic Uncertainty: Linguistic Invariances for Uncertainty Estimation in Natural Language Generation
International Conference on Learning Representations · 2023
510
cited
Towards Robust Evaluations of Continual Learning
arXiv.org · 2018
320
cited
Prioritized Training on Points that are Learnable, Worth Learning, and Not Yet Learnt
International Conference on Machine Learning · 2022
220
cited
Model evaluation for extreme risks
arXiv.org · 2023
198
cited
A Systematic Comparison of Bayesian Deep Learning Robustness in Diabetic Retinopathy Tasks
arXiv.org · 2019
121
cited
Uncertainty Baselines: Benchmarks for Uncertainty & Robustness in Deep Learning
arXiv.org · 2021
107
cited
Evaluating Frontier Models for Dangerous Capabilities
arXiv.org · 2024
103
cited
On Statistical Bias In Active Learning: How and When To Fix It
International Conference on Learning Representations · 2021
93
cited
A Unifying Bayesian View of Continual Learning
arXiv.org · 2019
76
cited
Active Testing: Sample-Efficient Model Evaluation
International Conference on Machine Learning · 2021
72
cited
Do Bayesian Neural Networks Need To Be Fully Stochastic?
International Conference on Artificial Intelligence and Statistics · 2022
72
cited
CLAM: Selective Clarification for Ambiguous Questions with Generative Language Models
2022
66
cited
Radial Bayesian Neural Networks: Beyond Discrete Support In Large-Scale Bayesian Deep Learning
International Conference on Artificial Intelligence and Statistics · 2019
66
cited
Prediction-Oriented Bayesian Active Learning
International Conference on Artificial Intelligence and Statistics · 2023
56
cited
Do Multilingual LLMs Think In English?
arXiv.org · 2025
55
cited
Prospect Pruning: Finding Trainable Weights at Initialization using Meta-Gradients
International Conference on Learning Representations · 2022
49
cited
Liberty or Depth: Deep Bayesian Neural Nets Do Not Need Complex Weight Posterior Approximations
Neural Information Processing Systems · 2020
44
cited
Discovering Agents
Artificial Intelligence · 2022
38
cited
Sotabase