Why neural foundation models work, and what they might—and might not—teach us about the brain (opens in new tab)

These models can partly generalize across species, brain regions and tasks, suggesting that a set of machine-learnable rules govern neural population activity. But will we be able to understand them?