I work at the intersection of psychology, philosophy of mind, and machine learning — applying what we know about how minds actually work to how we build AI systems.

Research Programme

My research has narrowed to a single thesis: Epistemic Developmental AI. The field builds AI by scaling statistical pattern matching; I ask what happens when you build it the way psychology and philosophy tell us minds actually work instead.

Curiosity is the hinge. A system driven by curiosity rather than a goal, built developmentally — grounded in concrete concepts, ordered by conceptual dependency, growing its own capacity — should recover the structure that generates language and reasoning rather than its surface statistics. And the same goallessness that makes it learn this way makes it safe: a system that only seeks to understand has no substrate for misalignment. It cannot deceive, because there is nothing it is trying to bring about.

For the full argument — the learning face and the safety face — see my research.

Background

  • AI Safety Researcher — Epistemic Developmental AI
  • Teaching Fellow, Computer Science, University of Warwick
  • Independent researcher

Contact

This site shares research progress, pilot studies, and writing on AI. All work is pre-publication and subject to revision.