Lee Sharkey

Research uncovers ways to interpret complex neural models using sparse autoencoders.

Access levels in AI audits influence effectiveness and risk identification.

A new approach improves feature learning in neural networks for better performance.