Lewis Hammond
Lewis Hammond
Room
012,
Wolfson Building,
Parks Road, Oxford OX1 3QD
United Kingdom
Interests
My DPhil concerns safety and incentives in multi-agent and machine learning systems, with a view towards building provably safe and beneficial AI, and in practice spans game theory, machine learning, and control and verification. My research interests are broad but are largely driven by the aim of improving robustness, explainability, and safety in AI systems, often through methods that combine both statistical and symbolic AI. Currently my efforts are focused on developing techniques to help rigorously identify or induce particular properties of multi-agent systems under their game-theoretic equilibria, especially those systems that operate in uncertain (partially known, partially observable, stochastic, etc.) environments. Related topics that I have worked on before include statistical relational learning, formal verification of learnt models, and deep symbolic reinforcement learning. Much of my work is also concerned with the problem of how agents represent and reason about preferences in the face of uncertainty. Examples of this include projects on representing non-Markovian reward structures using automata, learning models of cognitive biases using inverse reinforcement learning, and my master's dissertation on computational frameworks for moral decision-making. Thirdly and finally, I also take an interest in the governance, ethics, and societal impact of AI, though I do not research these topics full-time.
Biography
I am supervised by Michael Wooldridge, Alessandro Abate, and Julian Gutierrez, and am also a DPhil Affiliate at the Future of Humanity Institute. Before coming to Oxford I worked as an intern on Imandra and was a research assistant for Jacques Fleuriot at the University of Edinburgh, where I completed my MSc in artificial intelligence under the supervision of Vaishak Belle. Prior to this I studied for my BSc in mathematics and philosophy at the University of Warwick with Walter Dean.
Roles
- Future of Humanity InstituteDPhil Affiliate
Activities
- Probabilistic Model Checking
- Verification
- Machine Learning Systems
- Computational game theory
- Logics for strategic reasoning
- Complexity and equilibria
- Quantitative Analysis and Verification
- Rational verification