Anastasia Wei

I'm currently on leave from my Astrophysics PhD at UC Berkeley to work on AI safety. I think we're at a critical point, where highly capable AI systems could bring catastrophic and even existential risks relatively soon. I'm working with Second Look Research to replicate and stress-test empirical AI safety research, particularly on alignment training and character training techniques.

In my PhD and earlier research, I studied the physical origins of exploding stars in the night sky, searched modern astronomical surveys (wide-field sky surveys that image large portions of the sky repeatedly, enabling discovery of transient events, e.g. Rubin Observatory, La Silla Schmidt Southern Survey (LS4), and the Zwicky Transient Facility (ZTF)) for interesting targets (rare astrophysical transients: brief, luminous events that don't fit neatly into known categories, such as fast blue optical transients (FBOTs), tidal disruption events (TDEs), kilonovae, and peculiar types of supernovae), and built AI agents that automate science research workflows.

I’m always up for coffee (or tea, or boba) with new and old friends. If you’re passing through Berkeley, say hi!