´óÏ㽶ÊÓƵ

Please join us for a colloquium in Moore BO3 on Thursday, November 6, 2025, starting at 1:05 p.m., given by Wolford Lecturer, , Benjamin Scott Crocker Professor of Human Biology at Stanford University.

Title:  Bridging the Data Gap between Children and AI Models

Abstract:  Large language models show intriguing emergent behaviors, yet they receive at least three to four – and sometimes as much as six – orders of magnitude more language data than human children. What accounts for this vast difference in sample efficiency? I will describe steps towards a paradigm in which we can address this question. In particular, I'll discuss the use of egocentric video ("baby headcam") data for model training, and the use of developmental data for model evaluation. This paradigm provides a model-based framework for exploring the nature of children's early development. 

Coffee, tea, and cider donuts will be available a few minutes before and after the talk in the foyer space outside of Moore BO3.