Alexander Ku

I'm a PhD student at Princeton University (Psychology & Neuroscience), advised by Tom Griffiths and Jon Cohen, and a part-time Research Scientist at Google DeepMind.

I study how neural networks develop internal representations that support both automatic and controlled processing. My work focuses on the tradeoff between efficiency and flexibility—how systems can rapidly bind information in a controlled, context-sensitive manner while also forming stable, generalizable structures through automatic, statistical learning. I investigate how this tradeoff is optimized across different timescales, using tools from Bayesian statistics and information theory.

Email / Google Scholar / GitHub / Twitter / CV