Bounded Regret

Bounded Regret is the personal research blog of Jacob Steinhardt, hosted on Ghost at https://bounded-regret.ghost.io/. The blog originated on WordPress (with posts dating back to approximately 2010) and migrated to its current Ghost platform in October 2021. It is not a formal organization, nonprofit, or research institution — it is a personal writing platform operated solely by Steinhardt. Jacob Steinhardt is an Associate Professor in the Department of Statistics and EECS at UC Berkeley, where he is affiliated with the Berkeley AI Research (BAIR) lab and CLIMB. His academic work focuses on ensuring machine learning systems are understood by and aligned with humans, spanning topics such as neural network interpretability, reward misspecification, emergent behaviors in large models, and forecasting AI capabilities. The blog covers a wide range of topics including AI governance, AI safety research methodology, the philosophy of AI risk, empirical findings in machine learning, and forecasting. Notable posts include analyses of emergent deception and optimization in neural networks, examinations of complex systems and control, and discussions of how empirical findings in ML generalize. Steinhardt has also used the blog to publish call-for-proposals related to his work advising Open Philanthropy on AI safety grantmaking. Separately from the blog, Steinhardt co-founded Transluce in 2024 — a distinct nonprofit AI safety research lab focused on building open, scalable technology for understanding frontier AI systems. Transluce should not be confused with Bounded Regret, which remains his personal blog.