Free Energy Principle — Karl Friston

I was studying something similar in my MSc about the neurobiology of language. It's well known that reading circuits in the brain make predictions about which words come next - a sentence like "The jam on the motorway made traffic slow" doesn't elicit much activity, whereas "The jam on the motorway was sticky" does, because "motorway" makes it much more likely that "jam" was about slow traffic rather than the condiment. (There are perhaps some interesting parallels to LLMs there...) But we were looking in the visual cortex, way back before the brain is even perceiving words, and we were finding pre-activation of circuits even there depending on whether people were presented with real words (e.g. sheep) or non-words (e.g. brast). I think there's a lot to be said for conceptualising the brain as a statistical machine - it makes a lot of sense in evolutionary terms. There's a great video introduction to that idea with Karl Friston from UCL here: [this video]

This ties in quite nicely with Karl Friston's Free Energy Principle: [this video] There are so many strands of neuroscience that show predictive coding now. There's a really nice example in the use of common vs unusual words in sentences. If you're using fMRI to scan someone while they read these two sentences: The jam on the motorway made everything slow. The jam on the motorway made everything sticky. You'll see a flash of activity when people read the word "sticky" rather than the more common word in that context "slow". The theory is that when everything in the sentence proceeds as predicted there's very little need to update prior expectations, but when you come across something that goes against previous predictions it's a signal that something needs to be actively thought about.

Ok, so less about the flow and nature of information and more about the semantics. > Seeing the computation as a physical process, processing information "about" details in its environment, highlights this. Are you familiar with Karl Friston's work on the Free Energy Principle? [0] He touches on how the nature of a living being (computational agent) is to model its external environment and act upon that model. That making decisions based on internal logic requires external stimulus and semantics being analyzed and augmented in sort of a feedback loop. > The key to all this is appreciating that, once you see semantics and information processing as a matter of physical processes, you can see that the correct semantics can be necessary for producing a physical outcome. I think I understand what you're saying and it touches on some of my own inquiries. A more concrete example, if you have one, might better help align me with where your thoughts are at regarding this. [0] [this video]