Space is a latent sequence: A theory of hippocampus and PFC
Talkat Understanding Higher-Level Intelligence from AI, Psychology, and Neuroscience Perspectives · 2024-06-24..06-28·
summary
Claim is that space is a sequence and hippocampal phenomena arise from sequence learning via CSCG clone-structured cognitive graphs, an overcomplete HMM with hardwired emissions and learned transitions (EM/Baum‑Welch plus hard consolidation). Simulations in aliased 3D environments recover latent graphs, reproduce bimodal place fields after elongation, lap‑specific firing, remapping, and enable planning/navigation. Model adds fast contextual rebinding for in‑context learning and proposes CA3/CA1/dentate/PFC roles; implemented simulations match recent experimental analyses and practical navigation tests.
AI-written summary · about
chapters (11)
- 0:00Introduction to CSCG model
- 2:32Sequence learning mechanics
- 7:24The latent space of clones
- 16:23Predicting place field phenomena
- 27:16Geometric, lap, and remapping
- 37:24Model formalization and results
- 42:18Transfer and planning
- 47:16Cognitive schemas and composition
- 54:57In-context learning mechanics
- 1:01:31Microcircuit architecture
- 1:09:43Q&A and discussion
related talks
54mEmergent elasticity in the neural code for space
46mState-Space Models and Deep Learning: Is there a new revolution on our doorstep?
58mStrong Generalization from Small Brains and No Training Data
43mThe Biological Path Towards Strong AI
55mThe Prefrontal Cortex as a Meta-Reinforcement Learning System
56mDoes the Neocortex Use Grid Cell-Like Mechanisms to Learn the Structure of Objects?
55mComputation with Assemblies
53mAn Integrated Cognitive Architecture