The attendees of the workshop

The attendees of the workshop again

The first Agent Foundations for AI Alignment workshop took place in October 2023 at Wytham Abbey, Oxford, UK. Researchers from academia, industry and the wider community came together over a shared interest in the fundamental nature of goal-directed agency, and its application to the problem of ensuring current and future AI systems are safe and beneficial. The scheduled talks were a mixture of tutorials and original research.

Speakers

  • Tom Everitt (Google DeepMind)
  • Caspar Oesterheld (Carnegie Mellon University)
  • Abram Demski (Machine Intelligence Research Institute)
  • Vanessa Kosoy (Association for Long Term Existence and Resilience)
  • John S Wentworth (Independent Researcher)
  • Nisan Stiennon (Theiss Research/Encultured AI)
  • Michael Mandler (Royal Holloway, University of London)
  • Daniel Herrmann (University of Groningen)
  • Gerard Joseph Rothfus (University of North Carolina, Chapel Hill)

Organisers

  • Alexander Gietelink Oldenziel (University College London)
  • Nora Amman (PIBBSS)
  • Matt MacDermott (Imperial College London)
  • Kaarel Hänni (Caltech/Cadenza Labs)

Talks

The Jeffrey-Bolker Decision Framework (Daniel Hermann & Gerard Joseph Rothfus)


Regret Learners and Bounded Rational Inductive Agents (Caspar Oesterheld)


Generalisation and Agency from a Causal Perspective (Tom Everitt)


An Introduction to Reinforcement Learning Theory (Vanessa Kosoy)


Economics Without Preferences (Michael Mandler)


The Partition Assumption (Abram Demski)


Natural Latent Variables (John S Wentworth)