Predictive Coding: A theory of brain function that suggests the brain is always creating and updating a mental model of the environment. The brain actively predicts what sensory input will come next, rather than just passively registering it

https://chatgpt.com/c/6781f846-72d4-800c-b588-cd85abc6ef6a

• ANNs Hallucinations: predict there’s a pattern, and use analogous mapping to “guess” the blind spot. Aim is to minimize (cross) entropy

• ANN hypothesis: the detector takes in a pattern, and uses case based reasoning to fill in the rest (same with hallucinations)

  • Expm for Hypothesis: in the case of numbers, it would try to predict “the fourth one”. So in the framework, quantify this mapping and show how different inputs use it in analogous way.
    • measure this distance “that is closest” to show there is a numbers subspace

• BNNs:

  • https://medicalxpress.com/news/2011-04-eyes-brain.html
  • see frameworks for psychedelics: entropic brain theory, integrated information theory, and predictive processing.

• ANN hydra self repair

• BNN adaptation

  • https://www.scientificamerican.com/article/brain-adapts-in-a-blink/

Unifying Theories of Psychedelic Drug Effects

• Higher-level areas of the nervous system (i.e., higher-order cortical structures) generate top-down synaptic ‘predictions’ aimed at matching the expected bottom up synaptic activity at lower-level areas,
  • this multi-level hierarchical cascade of neural activity, high-level areas attempt to ‘explain’ the states of levels below via synaptic attempts to inhibit lower-level activity
  • lower levels will not ‘shut up’ until they receive top-down feedback (inference) signals that adequately fit (explain) the bottom-up (evidence) signals
• Predictive processing holds that our perceptions of external objects recruit the same synaptic pathways that enable our capacity for mental imagery, dreaming, and hallucination