Owen Parsons from Cohort 5 has recently spent some time at the University of Toronto and writes about his experience here:
“This past term, I was fortunate to spend four months at the University of Toronto, working with Dr. Roger Grosse’s research group. A core focus of the research group is to understand the training dynamics of neural networks e.g., why overparameterized models generalize well. My project focused on weak-to-strong generalization, a phenomenon relating to generalization in LLMs, by using properties of the Neural Tangent Kernel, a well-studied mathematical object in deep learning theory.
Beyond my research, I had the chance to engage with the broader academic community in Toronto. I attended seminars hosted by the Schwartz Reisman Institute, including one on “On the Origin of Rogue AI“. I also shared my work by giving a talk at Trajectory Labs, who host weekly meetups discussing the pursuit of a safe and beneficial AI future.
This exchange was a highly rewarding experience, both in terms of advancing my research and connecting with a collaborative and intellectually stimulating community.”
