Laws of Nature workshop: what can we tell from the program?

[Fra]

For comparison, here is the same talk delivered to the LoN workshop:
http://pirsa.org/10050056/
Law without law: entropic dynamics
Maybe my reaction is too one-sided. There may be valuable aspects of both presentations.

Last night I was able to watch to Ariels session.

From my point of view, Ariel describes quite well the core point of entropic inference and how it imples a flow of entropic time, and thus gives a kind of dynamics. There are good contact points with Verlinde as well. IMO this is an important ingredient in the big picture, and given that a lot of people doesn't seem to be tuned in on this reasoning at all: as was seen from the questions, people wondering what Q was.

But, apart from the basic idea of what is an entropic force, entropic dynamics and how does it yield expectations, there are many important points in the big picture that Ariel did not address at all. So I see Ariels talk as having a narrow focus, namely to present very briefly the reasoning and meaning of entropic inference, entropic dynamics and entropic forces.

The points where I have a very different view than Ariel is where he introduces QM.

In this talk you see that in the simplest case, in the way the reasoning is introduced, you just get a "simple" form of dynamics, which is basicall diffusion! So how can this inference model imply more complex dynamics, such as oscillatory phenomena or QM?

The assumptions QM uses in order to "imply" QM is not in my taste. What's good is that he illsustrates without much explanation a possible general mechanism, but I still think the connection is deeeper. I've seen several of ariels papers where QM is implied from various assumptions + inference logic, but the assumptions aren't justified IMO.

Also, another point which I think Ariels reasoning is not complete, is that his idea of the evolving statistical manifold is that the prior is updated, the manifold is updated. I think this "mechanism" is sort of right, but there are other mechanisms that I think is lost in ARiels and Jaynes reasonings since they start by assuming the the relative entropy measure that defiens hte measure on the manifold give the prior is unique. But I find the assumptions that go in there weak and ad hoc.

Instead I think the entropy measure itself is result of evolution.

So I disagree with quite a bit of Ariel says, buy I think his main message in the LON session here was to illustrate the most BASIC parts of what entropic dynamics is and how the general connection between dynamics on a statistical manifold, and the dynamics of the statistical manifold is connected. I share that, but to make sense of this and get the physics connection you end up questioning a lot of what Ariels is putting in as premises. ARiel doesn't seem to deny that though.

I *think* that it could be easy for some people to reject the reasoning since they don't see the big picture. Ariel does not line out the entire picture in that talk IMO.

I'm not sure how the other version of the talk Marcus referred to is different.

/Fredrik

 

[Fra]

Also, another point which I think Ariels reasoning is not complete, is that his idea of the evolving statistical manifold is that the prior is updated, the manifold is updated. I think this "mechanism" is sort of right, but there are other mechanisms that I think is lost in ARiels and Jaynes reasonings since they start by assuming the the relative entropy measure that defiens hte measure on the manifold give the prior is unique. But I find the assumptions that go in there weak and ad hoc.

Instead I think the entropy measure itself is result of evolution.

At first when Ariel argues that the manifold defined by the prior is fixed and when the prior is, and this is relaxed it sounds like if he has removed all background dependence, but he has not, because the entropic measure that defines the "set of manifold" of the transformation that generates the new manifolds is FIXED, this is still a background information.

This is something I object to and this is also evolving as I see it. Beucase the assumptiosn that goes into defining the equiprobability hypothesis and the space of probabilities based on continuuum probabilities is just too big and is IMO unphysical. From the point of view of pure inference and probability this is nothing strange, but I suggest that the connection to physics starts already here. Apparently Ariel does not think so.

Instead of adding uncelar constraints such as conservation of something we call energy, one could instead argue for conservation of complexity where the complexity is needed to encode all structurs. And no finite complexity can encode an infinite infinitedimensional space of continuum probability. This is one of my strong objections. This is something ariel probably didn't think a lot about since it traces all the way back to E.T Jaynes. But as I've rambled many times before on here I think the problem is in the first part of JAynes book, where he simply ASSUMES an apparently innocent (but I claim it's not) thing that "degree of beleif" is represented by a real number.

Surely, even rational numbers are real. but that's not the problem. The problem is not that degrees of believes are NOT real numbers, the real problem is the key property that you use in entropic inference, namely that the SET of all POSSIBLE degrees of beliefes suddenly is uncountable! For me that makes no physical sense!

/Fredrik

 

[Fra]

I listened to Philip Goyal's talk since the title was interested and Marcus mentioned it. I didn't actually watch the video or slides, I just used the mp3 while laying on a spike mat.

The general connection between classical logic, quantum logic and the corresponding properties of the real and complex numbers reminds me of some of the connections ariel has done in some papers where you can "dervice" QM formalism, by assumping properties of the complex number. This is not too unlike Jaynes, "derivation" of probability theory as an extension of inference but were you use the assumption the degrees of beleif is represented by real numbers.

But as I see it, while the connection is there (when you postulate that degrees of beliefs are represented as per aq particular number system, you do get a particular inferece logic - but what chooses the choice of mathematics in the first place?), the interesting problems were just metioned in the question part in the end. Royal seems to prefer the continuum because he is more familiar with it. I think what someone suggested, to make this "reconstruction" from rational numbers would be interesting.

He also said it's not clear to hime why the complex numbers and their properties seem to be prefered. I personally expect that question (why certain number systems are more FIT) is better addressed if reconstruct also the continuum rather than assume it.

/Fredrik

 

[ccdantas]

(...)

I still have to watch the video, no time yet.

Finally, only now I was able to watch this video.

Very interesting. Lots of material to think about, and potential points of convergence with other ideas. I am curious on how far the idea of time as a "change of change" is connected with H. Bergson's idea of "duration". Hopefully, the book will be released soon. It might have some reference to Bergson.

For those who have not watched it yet, it's not a waste of... time.