Cognitive Scientists' Criticism of Karl Friston's Free Energy Principle

Prof. Paul Thagard criticises Karl Friston's renowned Free Energy Principle, claiming it's nothing more than predictive processing, which is an insufficient theoretical framework for the mind.
Watch the full podcast here: • Paul Thagard on AI, Fr...
---------------------------------------
{Podcast}
Substack: rsampod.substack.com/
Spotify: open.spotify.com/show/4ryEqju...
Anchor: anchor.fm/rahul-samaranayake
{Website}
rahulsam.me/
{Social Media}
/ trsam97
/ name_is_rahul
/ rahul-samaranayake-981...
---------------------------------------
If the ideas I discuss in this channel evoke your interest, consider visiting theunhappyman.substack.com/
---------------------------------------
Copyright Disclaimer under section 107 of the Copyright Act 1976, allowance is made for “fair use” for purposes such as criticism, comment, news reporting, teaching, scholarship, education and research.
Fair use is a use permitted by copyright statutes that might otherwise be infringing.
If you are or represent the copyright owner of materials used in this video and have a problem with the use of the related material, please email me at [e-mail], and we can sort it out. Thank you.

[@qhansen123]

I felt the that the free energy principle wasn’t obscure but couldn’t find opinions against the idea until seeing this video.
to try and test predictive processing I used an RNN to classify MNIST images by glancing at patches, then added a side loss function of predicting the next patch. To test full predictive coding only the KL divergence between predicted and real input was passed as input, and it could still get above 80% accuracy. Similar results from using the difference and MAE as error signal. This was a 3am obsession and not for any real acedemic research.
What are the conclusive arguments against predictive coding?

[@therrealquickquack]

Variational free energy is a well-defined quantity used in variational inference, it's not an "analogy with thermodynamics free energy" at all (I mean, of course they used this name because it's something that is minimised I guess, but that's all), it's sad to see people who have no clue on certain things talking about them with such confidence.

[@RahulSam]

That’s true. In fact, I’m starting to learn more about its specificities, too, as I just started reading the book Active Inference.

[@sharkbaitquinnbarbossa3162]

Omg. According to Thagards wiki article and the full interview, he is supposed to be strongly influenced by pragmatism and Peirce. It's hilarious and shows how people from cognitive science often lack competence in basics of information theory and statistcs in general, which is nothing else than a realisation of abductive inference (basically since conditional probability / Bayes rule says nothing else in the form of hypothesis testing). Thagard is a very good example of typical hegemonic relativism and gate-keeping strategies in philosophy and science, a non-explanatory tirade that just covers opinion (relevance theory by Sperber / Wilson is of the same kind, to give another example). As a student I am more and more disgusted that a rather privileged crowd of academics are more occupied in performing a kind of habitus, for which they constantly asks for confirmation, instead of providing transparent argumentation... I am angry at this not because he is critizing Friston or so, but apparently he seems to actual have the tools to understand something he calls bogous (again, since he writes about abductive reasoning etc., which is what the FEP represents; same fallacy in relevance theory - mere gate-keeping, since his coherence concept appears to not differ much or at all conceptually from what active inference or the FEP says - that is what is bogous). The same attitude can be found in stats: people that can't properly define and explain p-values, but give you a lecture in p-hacking etc. Academics is rotting away in attitudes of pseudo-differentiation and habitus approval... I also like his desciption of Bayes and its boundaries, after talking about abduction, which in my opinion represents very well the problems which a human has, trying to perform inference via mere abductive inference (which at the end just means: every inference we make is probabilistic, as it just represents the inferential modality of contingency of inference, since we do not have absolute knowledge, where every inference can be represented as necessity modally (either 0 or 1, not in between))... St.Sch.

[@RahulSam]

Insightful comment and I agree with some of your criticisms of “academic habituses.” I’m hoping to talk with Karl Friston this year, and I’ll bring up some of your points.