Rory Greig

Rory Greig

I am a Research Engineer on the AGI Safety and Alignment team at Google DeepMind, working on Scalable Oversight techniques (e.g. Debate)

I have experience in machine learning, evolutionary computing, agent-based models, and high-performance simulation. I am interested in AI Safety, Complexity Science and Artificial Life (ALIFE)

Experience

Senior Research Engineer

Google DeepMind | January 2022 - Present

  • Working on techniques and infrastructure for both near term and longer term AGI safety.
  • Focus on aligning Large Language Models (LLMs) to human preferences using Reinforcement Learning finetuning.

Software Engineer and Research Scientist

Improbable | 2015 - 2021

  • Led research on genetic programming to learn symbolic logic for agent-based models from scratch.
  • Worked on GPU acceleration for high-performance COVID-19 agent-based models.

Selected Publications

  • "Improving alignment of dialogue agents via targeted human judgements" (September 2022)
  • "Generating Agent Based Models From Scratch With Genetic Programming" (ALIFE Conference, July 2021)
  • "A Dynamic Microsimulation Model for Epidemics" (Social Science & Medicine, December 2021)

Contact

Email: rorygreig@gmail.com