Ryan is interested in building theory for safe AI design. Recently, this has focused on analyzing agents’ incentives using causal frameworks. Previously, he has also analyzed alternatives to utility maximization, such as value learning and quantilization. Before that, he worked as a medical doctor, and has degrees in medicine and bioinformatics from Monash University and Imperial College London.
- The Incentives that Shape Behaviour. (Ryan Carey, Eric Langlois, Tom Everitt, and Shane Legg, SafeAI@AAAI)
- How useful is quantilization for mitigating specification-gaming? Ryan Carey. (SafeML ICLR 2019 Workshop)
- (When) Is Truth-telling favored in AI Debate? (Vojtech Kovarik, Ryan Carey; SafeAI@AAAI)
- Predicting Human Deliberative Judgments with Machine Learning. (Evans O, Stuhlmüller A, Cundy C, Carey R, Kenton Z, McGrath T & Schreiber A, 2018)
- Incorrigibility in the CIRL Framework (Ryan Carey; Proceedings of AI, Ethics and Society)