I’m a PhD candidate in the engineering department at the University of Cambridge. I’m interested in solving problems that maximise expected human prosperity, and I’m generally method-agnostic. In the past, that meant working on technical problems related to climate change mitigation, but I now believe that the innovation required there is political not technical. More recently, it’s meant working on problems in reinforcement learning. It seems clear to me that building intelligent agents that are safely aligned with human values would provide huge social benefit. So my PhD focuses on doing just that.

Current areas of interest include:

  • Zero-shot RL
    • pre-training RL agents capable of solving any task in a given environment;
  • Open-ended learning
    • designing environments and agents that co-evolve for unending generality.

I don’t believe many things strongly, but at the time of writing (April 2024) I’m willing to defend the following:

  • Progress is possible and desirable.
    • progress means solving problems.
    • and all problems are soluble in the right knowledge solves them, we just need to find it.
  • The goal in life is to maximise your cumulative happiness.
    • whilst (kind) people may help you achieve this, you cannot expect them to, it is your responsibility alone.
    • pursuing this (perhaps unintuitively) greatly benefits others i.e. it’s a positive sum game.

If you are interested in any of the above, I’m always keen to chat so send me an email (srj38 (at) cam(dot)ac(dot)uk) or hit me up on twitter/X @enjeeneer.