Whether or not anything "breakthrough" comes out of his work, I'm very sympathetic to the kinds of explorations he does ... especially since new things go through a messy phase before clarity arises.
> I'm very sympathetic to the kinds of explorations he does
Yes, and I think there is a success criterion that most people are not considering, which is: success as a less complicated explanation for the same phenomenon.
When people ask for new testable predictions as the only way a new theory can be successful, they are revealing that they don't consider parsimonious explanations to be a scientific goal.
Stephen's work with hypergraphs shows a lot of promise as a simple theory that implies other successful theories of physics. It focuses on emergent phenomenon that exhibit "pockets of computational reproducibility". The behavior of those pockets can then be predicted with tools in the physics canon like Riemannian geometry or complex numbers.
> When people ask for new testable predictions as the only way a new theory can be successful, they are revealing that they don't consider parsimonious explanations to be a scientific goal.
It doesn't have to be new testable predictions, but it should at least reproduce some testable predictions of the conventional theories.
As far as I know, these discrete graph models have been shown to reproduce some broad features of QM and GR (mostly the work of Jonathan Gorard rather than Wolfram), but they don't make any actual numerical predictions, which in my view is a basic requirement of a physical theory.
At the moment it seems like some intriguing toy models, similar in many respects to t'Hooft's cellular automaton model of QM, but there's no physical theory here
True, and I always appreciate how he built the wolfram language to allow himself these kinds of explorations, which would be next to impossible without it.
