Stephen Wolfram solves physics.

[halfeye]

https://writings.stephenwolfram.com/...its-beautiful/

Anybody know enough to say this is or isn't hogwash? If it's legit it sounds very good, and Wolfram is known for being clever.

[NichG]

This is more of a framing than a solution. It seems to be related to things like lambda-calculus, graph rewriting grammars, and the like, and in that sense it isn't wildly new. Though this is a more general form than those things, and the study of the distribution of the different kinds of outcomes you get by exhaustively enumerating the rules is very much Wolfram's style (he did the same thing for 1d 2-color neighborhood-1 CAs in New Kind of Science).

What remains to be seen is whether or not using this kind of framing for a physical process or system actually lets you calculate, predict, derive, or construct things that would be more difficult to get at in other ways. For physics types of problems, it doesn't pop out at me why I'd want to represent the problem this way rather than with fields, operators, or other formalisms we already have - though this doesn't look that distant from stuff like Feynmann diagrams, so maybe iterating the graph rewrite rules a few times lets you easily generate all the combinatoric terms of some particular useful expansions. But for chemistry problems, especially things like open-ended chemical spaces, it seems like it could be useful.

[Khedrac]

I have no idea if he's really on to anything, however the beginning of what he is discussing is basic complexity theory - complex behaviours emerging from simple rules - though he seems to be concentrating of a form where simple behaviours emerge from the complex behaviours that the original simple behaviours created.

Even if it cannot do what he wants (explain the universe) it probably will provide a powerful tool for mathematicians (and all scientists that use mathematics) for the future.

[Rakaydos]

I have no idea if he's really on to anything, however the beginning of what he is discussing is basic complexity theory - complex behaviours emerging from simple rules - though he seems to be concentrating of a form where simple behaviours emerge from the complex behaviours that the original simple behaviours created.

Even if it cannot do what he wants (explain the universe) it probably will provide a powerful tool for mathematicians (and all scientists that use mathematics) for the future.

To be fair, we know quantum physics is a much more complex behavior than general relativity, and any grand unifying theory of physics must explain the bigger (relativity) using the outcomes of the smaller. (quanta)

[halfeye]

To be fair, we know quantum physics is a much more complex behavior than general relativity, and any grand unifying theory of physics must explain the bigger (relativity) using the outcomes of the smaller. (quanta)

That reads as if you didn't follow the link. The text is quite long, but it's probably worth reading IMVHO.

<edit>

He seems to say black holes drop out of normal space-time and become "frozen stars", but I don't believe he wrote about Hawking Radiation, does anyone know whether he's saying there is no such thing?

[NichG]

That reads as if you didn't follow the link. The text is quite long, but it's probably worth reading IMVHO.

<edit>

He seems to say black holes drop out of normal space-time and become "frozen stars", but I don't believe he wrote about Hawking Radiation, does anyone know whether he's saying there is no such thing?

It looks more like he's saying 'look, on some directed graphs there are regions you can enter but not leave ; some of these rules produce graphs with such regions that occur at branching points; those are equivalent to event horizons in the model, and occur naturally from the dynamics of construction ; therefore, this type of model could express black holes'

Edit: okay, it's slightly different than that since the directed graph he's talking about there isn't the hypergraph itself, its the graph of relationships between the update events (what he calls the causal graph). So its a bit more nontrivial.

Almost everything in this isn't making declarative statements about our universe, it's showing that the graph rewriting framework is general enough to contain rules to express various phenomena similar to ones in our current theories of physics.

With this kind of approach, you can't predict if our universe has Hawking radiation, you can only ask whether something like Hawking radiation can be expressed or not.

Edit2: the result that gets the Einstein equations from the constraint that the graph has an asymptotically constant fractal dimension is pretty cool!

[tiornys]

I'm very much an armchair amateur physicist, but I think what he's trying to do is present as abstract a form of modeling as possible given the constraint that we want to unify physics. In much the same way that String Theory or Quantum Loop Gravity aren't so much single theories as classes of models that have different behaviors depending on inputs, this is also a class of models that have different behaviors depending on inputs. It looks to me like this has less inputs that are more abstract than any of the other major unification theories I've looked at.

So if I've gotten all of that correct, then this looks to me like a generalized version of unification attempts that should in principle be capable of modeling all of those unification attempts in addition to others we haven't tried yet. If that's actually the case, and if we can search this model space at least as efficiently as we can search String Theory model space/Quantum Loop Gravity model space/etc., then this really is a revolutionary approach.

Or, it might just be an overly hyped, differently presented version of stuff we've already examined. I don't know enough to say.

[halfeye]

Edit2: the result that gets the Einstein equations from the constraint that the graph has an asymptotically constant fractal dimension is pretty cool!

The whole thing is immensely cool, if it works he actually has almost solved physics. I don't know whether he's a nice guy (which is not in any way to imply he's not), but he's already done interesting things before this.

I wonder what Douglas Hofstadter makes of it, because he was into that sort rules for string manipulation thing once, I wonder whether he is still interested.

I'm very much an armchair amateur physicist, but I think what he's trying to do is present as abstract a form of modeling as possible given the constraint that we want to unify physics.

That's not what he says, he was messing about with some pretty patterns and by chance noticed it looked a lot like physics, and then another way up it looked like some more, different, physics; so there was a mapping between the two aspects that resolved the problems between quantum theory and relativity.

[Chronos]

Having a framework that can be used to describe physics isn't the same thing as solving physics. I mean, you can also describe physics using equations, but that doesn't mean that once you've taken algebra, you've solved physics.

What he did might be useful and it's almost certainly interesting, but it's not the be-all and end-all.

[NichG]

The whole thing is immensely cool, if it works he actually has almost solved physics. I don't know whether he's a nice guy (which is not in any way to imply he's not), but he's already done interesting things before this.

I wonder what Douglas Hofstadter makes of it, because he was into that sort rules for string manipulation thing once, I wonder whether he is still interested.

One thing that occurs to me is that to get those results, he had to go to this 'causal invariance'/causal graphs framework, and when I first saw it I thought those causal graphs were the same as the graphs his rules rewrite. But they're actually something else that can be derived from the rewrite rules - but could also apply to systems that aren't graph rewriting systems. So I wonder how much the insights about foliations of causal graphs, causal invariance, edge flux crossing sets of surfaces corresponding to conserved quantities like energy and momentum, etc actually don't depend on the rule stuff at all.

E.g. if I squint at it, the conditions he gives for causal invariance are basically ergodicity and the existence of a unique equilibrium. So that's actually quite general - a lot of different frameworks can have those properties. Could we see Einstein's equations fall out of those systems too, in the same way?

If that's the case I'd be tempted to say that this is really two separate works, and that the observation that 'any system which asymptotically behaves like a space whose dimension does not depend on where or when you are obeys general relativity' is probably a bigger result than the specific stuff about graph rewriting rules.

[aspi]

I'm going to be honest: I lost interest halfway through the introduction and scrolled through since Stephen Wolfram is notorious for this kind of approach. The fact that the term "grammar" occurs exactly zero times in the entire articles makes me feel like he is just reinventing the wheel without regard for prior work and is simply phrasing (some) physics in terms of generative type-0 grammars and then playing around with that notion without putting forward any actionable hypotheses.

So a question for those of you who did read it properly - am I missing something with that assessment that would make this worth the ~1 hour to really read it?

[137beth]

I'm not a physicist: I'm a mathematician. So I can't really evaluate his most recent claim. However, I will state that Wolfram has a long track record of using self-published writings to claim other people's discoveries as his own, as well as to exaggerate the significance of his own discoveries.

In 2002, Wolfram self-published a book called A New Kind of Science, which covers a range of topics in discrete dynamical systems, an area of math which is rarely taught in classes at any level of our school system. I read the book, and I was impressed by the breadth of topics he discussed, but it had several problems that would have been caught during peer review if he hadn't self-published. I could describe those issues at length, but someone else already did it better than me: Lawrence Gray's article "A Mathematician Looks at Wolfram’s New Kind of Science", contains most of the criticism I have for Wolfram's book, and Gray explains it better than I could.

The short version, though, for those who don't want to read the whole article, is as follows.

1)Most of the information in Wolfram's book was previously published by other authors in other sources. That on its own isn't unusual: in math, new discoveries are usually published in papers, and books collect previously-published information in a form which is accessible to someone learning the subject. However, Wolfram repeatedly claims in his book that those discoveries are his, even though they aren't.

2)The book contains one important theorem which had not previously been published. It states that a particular dynamical system (which Wolfram calls W110) is Turing-complete (i.e., it can simulate any computer program). This is an important result which Wolfram claims credit for, but it was actually proven by Matthew Cook, who at the time was a student interning at Wolfram Research.

3)Additionally, the book doesn't actually contain the full proof of Cook's theorem. It contains a summary which brushes over important details. That kind of summary is appropriate for an oral presentation, or some other context where time and space are limited. However, a summary of a proof should always include a reference to where the reader can find the whole proof, and NKS doesn't provide such a reference. Also, the book is over 1000 pages long, so he could have fit the complete proof in it. Wolfram also sued Cook in an effort to prevent him from releasing the full proof before the book was published.

N.B.: In 2004, Cook released a 40 page paper detailing the proof that W110 is Turing-complete. Cook also describes the motivations for how he thought of each step (something which, unfortunately, is often left out of math papers).

4)The big qualitative idea that Wolfram repeats over and over in NKS is that a dynamical system with very simple rules can produce complex behavior, and in particular can be Turing-complete. He claims that this idea is his own, and cites "his" theorem about W110 as evidence for the principle's validity.

However, this same idea is expressed 20 years earlier in Conway and Guy's book Winning Way for your Mathematical Plays (1982). In Winning Ways, the authors prove that the "Game of Life," a dynamical system with simple rules, is Turing-complete. Once again, Wolfram claimed someone else's ideas as his own.

N.B.: Conway recently passed away from COVID-19.


Anyhow, the main point I am trying to get across is that if Stephen Wolfram claims to have discovered something in a self-published blog post, I am very skeptical both of the significance he claims the discovery has, and of whether or not Wolfram actually made the discovery himself. His past record of academic fraud has given me plenty of reason to be skeptical.

[TaiLiu]

“There’s a tradition of scientists approaching senility to come up with grand, improbable theories,” the late physicist Freeman Dyson told Newsweek back in 2002. “Wolfram is unusual in that he’s doing this in his 40s.”