I currently work on model personality and post-training research at OpenAI. Humanity is building minds of enormous complexity. Understanding and shaping their character feels to me like one of the most important tasks of our time. I’m particularly interested in good evaluations and how to learn scalably from human oversight. Please let me know if you have thoughts on GPT’s personality!
Before this, I dropped out of an ML PhD at CMU and did my undergrad at Cornell. My research includes AidanBench (opens in a new tab), Representation Engineering (opens in a new tab), and Localizing Lying in Llama (opens in a new tab).
As a hobbyist historian of AGI, I also created AI-Timeline.org (opens in a new tab).