Cambridge, Massachusetts, United States
Research Resident on the Alignment team.
Wrote a piece on the implications "elicitation" has on AI evaluations.
Provided feedback to Anthropic’s models on their alignment research outputs.
Helped write initial drafts of the AISI's Managing Misuse Risk for Dual-Use Foundation Models: https://nvlpubs.nist.gov/nistpubs/ai/NIST.AI.800-1.ipd.pdf. Wrote policy memos. Provided technical feedback.
Research position, mentored by Evan Hubinger. My research is on getting models to accurately self-report internal states, for example, accurately reporting the presence of a steering vector.