Introduction E9F5FC

Questions FFFFC0

• View
• Edit
• History
• Print

Biology

Active Inference

• First mind: minimizes "surprise"
• Second mind: distinguishes sadness and surprise and minimizes sadness but tests and explores its models in areas that are not so important, the theoretical world, thus allowing more room for surprise
• Third mind: minimizes anxiety - keeps the second mind from gaming expectations, from wishing what it does not desire

Active inference explains how people are inclined to confirm their biases. The purpose of consciousness is to force people to confront their biases, to investigate, to take a sober, honest look, which way is it.

From Chris Field. Physics is Information Processing. Lecture 1. Entropy only makes sense as a discrete concept. This reminds me of how entropy is calculated by carving up phase space with a discrete grid, which implies there is an observer imposing a specific coordinate system. Also, the relationship with symmetry, where there are different actions which may yield the same effects.

• Consider a boundary across which discrete information is exchanged. The boundary exists in phase space, not in physical space. Not a theory of motion, but a theory of information transfer. Consider questions with yes-no answers.
  • An action that asks a question, a thing/being to whom the question is addressed, a shared language to ask the question in.
  • It takes time and energy to ask a question. Action = {$E\times t$}. Minimum action to ask a question is {$\hbar$}.
  • Information always has to have some encoding. That encoding is physical. It can be called the reference frame.

Bayesian inference involves thinking backwards and as such is relevant for the fivesome.

https://en.wikipedia.org/wiki/Holographic_principle Relationship between holographic principle, Markov blanket and free energy principle.

Chris Fields has a diagram of a quantum reference frame with two chains with arrows in both directions, the measurement chain and the preparation chain, and with arrows passing from locations in the measurement chain to the outcome classifier and then to locations on the preparation chain. Is this related to the fivesome, sixsome or sevensome? The skeleton of an autoencoder.

Chris Fields. A reference frame allows for measurement but requires and active participant to do that measurement. And the result is approximate, especially when communicated.

Kullback-Leiber divergence: Principle of minimum discrimination information

• Compare minimizing anxiety (not have unhealthy expectations) and minimizing sadness (risking unhealthy expectations) and minimizing surprise (active inference).
• How is minimization of anxiety different from minimization of surprise or sadness?
• How is existential readiness, preparedness different from minimization of surprise?
• How is minimization of free energy related to minimization of surprise?
• Temperature allows for a spectrum of equilibriums for reactions, which can support a range of layers of randomness and regularity.
• Chemistry - spontaneous reactions - proceed in one direction (assembling) or the other direction (breaking down) - depend on whether the temperature is low or high - and that sets up an adjunction between the two sides as with free (construction) and forgetful (losing structure). Determined by change in Gibbs free energy {$\Delta H - TS$}, whether it it positive or negative.

Free energy = slack

Consciousness requires the consideration of two different kinds of Bayesian priors - a presumption of knowing and a presumption of not knowing.

Consider the varieties of energy: External (global) kinectic and potential. Internal kinectic and potential. Work and free energy. Sixsome? What more?

• Active Inference. Grandparents impacts parents, parents impact chldren, but also, can grandparents impact their grandchildren directly? (As with orthogonal polynomials?)
• Is the Active Inference loop related to the semiotic square? Negation determines, distinguishes active states and sensory states.
• Study how Judea Pearl uses a Markov blanket.
• Active Inference. How can it distinguish surprise and sadness?

inference = "way of figuring things out"

Structure Active Inference Generative model creates beliefs, the policy creates actions, but our actions are "fated". We are only "free" to choose our beliefs. So our consciousness has to do with the beliefs we take, our will. We don't have to experience our Bayesian inference, we can simply identify with it, take ownership of it.

• Toby StClere is interested in understanding action better.
• Looking for a canonical way to choose action or do inference. This sounds like setting up a free functor, which means defining what you want to forget. Related to the maximum entropy principle.
• Consciousness is a forgetful functor (forgetting experience and language) with a free functor.

Karl Friston - your universe is trying to infer you (given that you are trying to infer your universe).

Orthogonal polynomials are based on a recurrence relation which is second order thus not a Markov chain, not modeled by Active Inference.

Spirtes, Glymour and Scheines. Causation, Prediction and Search. (Adaptive Computation and Machine Learning).

Modeling experience from old self to new self - Bayesian analysis - prior belief + new evidence = revised belief