You can find the white paper The Free Energy Principle Made Simple But Not Too Simple.
Further discussion included links on the following topics:
Also, here are some notes from the session by Nathaniel Evry from Quub:
My General notes reconciling the Free energy principle and hippocampal formation as a modality-agnostic mapping machine and the notion of connectedness.
- The hippocampus can be seen as a set of internal states that are separated from external states by a Markov blanket of sensory and active states. This fits with the hippocampus integrating multimodal sensory inputs and influencing memory encoding/retrieval.
- Sparse coupling in the hippocampus could give rise to its hypothesized property of “separated comparators” – where different input streams are kept independent for comparison. This is analogous to conditional independencies between internal states.
- The hippocampus is thought to create a coherent spatial representation through attractor dynamics. This fits with the paper’s notion of a steady-state density implying an attractor. The coherent representation could be seen as a low entropy steady-state.
- Mapping spaces and the notion of connectedness relies on the hippocampus integrating information across modalities into a coherent whole. This is analogous to internal states synchronizing with external states through active inference.
- The principle of least action suggests hippocampal dynamics will follow flows that minimize “surprisal” or prediction error. This connects statistical properties with self-organization and could account for phenomena like memory consolidation.