I am interested in building intelligent systems, using machine learning. Firstly, there is the challenge of trying to make computers capable of learning about the world first hand from their own data, instead of being told what to do by direct instructions from humans. And secondly, there is the continuing mystery of how real brains achieve the same thing, with fantastic success, despite being made of fallible, slow processing elements. So a longer term goal of my work is to shed light on how the brain comes to represent and predict the world, and how it uses this knowledge to generate sensible actions. Within machine learning, my current interests are inference and optimisation in neural nets, belief nets, and Gaussian processes.

A second research direction is complex adaptive systems and evolution. I have studied the evolution of cooperation (especially the "tragedy of the commons"), and the strange effects generated by cyclic competitions ("rock-paper-scissors") between species. Recently, I have looked at the way that network structure in the interactions between creatures can affect their evolution.

One thing that binds these interests together is the emergent behaviour of communities:

• how do collections of dumb neurons add up to intelligent brains?
• how does the web of who-meets-who affect the direction taken by evolution?
• how do cooperative entities emerge from dynamical processes that are "red in tooth and claw"?