Herbie Bradley
Herbie Bradley is a PhD student in the Department of Engineering at the University of Cambridge, researching synthetic data generation and open-endedness in large language models. He previously served as a technical staff member at the UK AI Security Institute and Frontier AI Taskforce from 2023 to 2024, where he worked on model evaluations and organized the Bletchley Summit. He is a co-author of the Pythia suite for analyzing large language models and has contributed to EleutherAI and CarperAI research projects. He has published at conferences including ICML, NeurIPS, ICLR, and FAccT.
“Herbie Bradley asked, currently this level of potential misalignment basically prevents deployment or if deployed would prevent further deployment if an incident happened in a customer's deployment.”
Source→AI-extracted from podcast / newsletter / paper summaries. May contain errors.