On Fri, Jun 28, 2019 at 9:40 AM Matthew Brown <[email protected]> wrote:

> https://arxiv.org/abs/1906.10184
>

A free energy principle for a particular physics
Karl Friston

Nice find. Friston going quite far down the "subjective world" rabbit hole.
I quite often draw this parallel, so this paper may be useful to me.

It's nice he sees the parallels. But to me he is just fitting a formalism
to an underlying observation that cognition and our perception of physics
have parallels. Cognition behaves in many ways like quantum mechanics. It's
been observed before. Though it is true that this is not yet embraced or
understood, and I would identify that as the key problem holding AI back.
This is just another formalism to express that. There are others. To me the
underlying insight is just the surprising complexity you get from emergent
structure.

However, if Friston brings more focus to this, it may be the path for the
field to understand and embrace it. Which will be good.

Though I think the essential insights can get buried in his formalism.
Crucially I think non-linearity of combinations gets lost (in his
generative models.) For instance I'm not sure if the linear simplification
of Markov blankets is justified. Assuming relationships are linear so the
maths can be tractable strikes me as a key failing in most Active Inference
work.

-Rob

------------------------------------------
Artificial General Intelligence List: AGI
Permalink: 
https://agi.topicbox.com/groups/agi/T903bb1e5785b431f-M0e39373793808d2012b0328e
Delivery options: https://agi.topicbox.com/groups/agi/subscription

Reply via email to