This paper pokes at one of the central parts of HTM theory.
https://www.sciencedirect.com/science/article/abs/pii/S0149763423003731?via%3Dihub
2 Likes
Based on the abstract and intro (I don’t have access to the rest), I’m concerned that the paper is mixing up a couple things:
- “Falsifiable” versus “Falsifiable using EEG / MEG”
- “Predictive coding is one of the things that the thalamocortical system does” versus “Predictive coding is the only thing the thalamocortical system does” versus “Predictive coding is a grand unified theory that explains everything in the entire brain”.
They do seem to talk about the second one, or at least something close to that:
In this way, we seek to differentiate between two claims: 1) predictive coding explains a large part of the behaviour of the brain; and 2) predictive coding explains all of the behaviour of the brain. A positive response to the first of these seems difficult to argue against – there is a substantial extent to which the brain seeks to predict the world. This paper specifically considers whether the second of these claims is supported, or at least lays a foundation for how to empirically test it.
Anyway, my take is that predictive processing is one of the main things that the thalamocortical system does, predictive coding (in the traditional algorithmic sense) is much more specific than that and not centrally relevant to the brain (but probably shows up here and there), and the Friston-style “this is a grand unified theory of the whole brain, including each and every random obscure nucleus in the brainstem” is just baloney. More specifics here:
3 Likes
Damian Cruse
@damiancruse
Hi. The open access side is taking its time, so in the meantime you can download the accepted manuscript here: https://damiancruse.com/_files/ugd/38b
2 Likes