.

And by a prudent flight and cunning save A life which valour could not, from the grave. A better buckler I can soon regain, But who can get another life again? Archilochus

Friday, January 24, 2025

The 'Free Energy' Universal Brain Theory

The Learning Model behind Generative AI
Generative AI Dream States
The first line of "Finnegan's Wake" is "riverrun, past Eve and Adam's, from swerve of shore to bend of bay, brings us by a commodius vicus of recirculation back to Howth Castle and Environs" and the last line is "a way a lone a last a loved a long the" - essentially ending mid-sentence, which loops back to the beginning of the book to complete the sentence.

The entire work forms a cycle, the book ending with the sentence-fragment "A way a lone a last a loved a long the" and beginning by finishing that sentence: "riverrun, past Eve and Adam's, from swerve of shore to bend of bay, brings us by a commodius vicus of recirculation back to Howth Castle and Environs."
Ala - Schizoanalysis and Schreber
Perception vs Learning
Excerpt from 1st video:
...If the generative model predicts sensory patterns that don't match what you're actually experiencing, your brain adjusts the explanation and tries again. This back and forth continues until your brain finds an explanation that minimizes free energy, one that explains both the sensory input and aligns well with your prior beliefs. This is the process of perception, when the brain rapidly adjusts the activity of those latent neurons we talked about, tweaking its explanation until it finds the one that minimizes free energy. This happens incredibly quickly, usually within fractions of a second.

But there is also a longer term process of learning. Over time, the brain gradually refines both models by adjusting connection weights between the neurons. The recognition model gets better at making initial guesses, and the generative model gets better at predicting their sensory consequences and builds up better prior expectations of causes.

Importantly, even though perception and learning operate on two different time scales, they both serve the same overarching goal, reducing uncertainty in the world by building optimal models of the environment and finding explanations for sensory data within those models.

MacArthur's Park is melting in the dark, all the sweet green icing flowing down.

Someone left a cake out in the rain.  I don't think that I can take it, for it took so long to bake it, and I'll never have that recipe again, oh no!

No comments: