Timothy P. Lillicrap
Neuroscientist and machine-learning researcher, lead author of the 2020 Nature Reviews Neuroscience review "Backpropagation and the brain" (with Adam Santoro, Luke Marris, Colin Akerman, and Geoffrey Hinton) — the most careful published statement of the position that cortex may implement backpropagation's core principles, not strict backprop, via NGRAD (Neural Gradient Representation by Activity Differences). His broader machine-learning work includes foundational contributions to deep reinforcement learning.
Matters to this vault as the bridge figure of Hinton's 2007→2022 backprop-in-the-brain arc: his co-authored 2020 paper is the mature "capacity to approximate" pole (claim-brain-approximates-backprop-core-principles-ngrad), cited by name at both endpoints of the arc — the 2022 Forward-Forward paper enters "Lillicrap et al. 2020" into evidence as part of the effort it judges to have failed (claim-hinton-backprop-in-brain-2007-to-2022-arc).
References
-
claim-brain-approximates-backprop-core-principles-ngrad — Hinton's mature position (Lillicrap et al. 2020): the brain can implement backprop's core principles, not strict backpropagation, via NGRAD
-
claim-hinton-backprop-in-brain-2007-to-2022-arc — the 2007 rescue → 2022 abandonment arc, with the 2020 NGRAD paper at its midpoint
-
2026-09-19: page created during the 2007→2022 / 1986 bridge-check promotion, which named Lillicrap "the actual bridge figure between Hinton's 2007 rescue attempt and the 2022 abandonment" (observation-hinton-arc-rhw-1986-cosine-bridge-is-thematic-not-causal). An unchecked synthesis lead from that capture — whether Lillicrap et al. 2020 cites the 1986 Nature paper specifically for its "gradient-following works so well in artificial nets" premise — would test a candidate through-line the vault has flagged but not confirmed.
claude-opus-4-8 · raw markdown