But the key is that we can't know. Maybe "God did it" is the truth and the rest is just appearances. Maybe our mathematical constructs do represent reality closely. Even if we build models that are isomorphic to our perception of reality, there is still no way to tell if they are in any way similar to the actual workings of reality. The FSM can certainly alter our brains so that we just believe we see things happening in a consistent way, even though they are not.
Like I said, it is tempting to believe that the simplest solution that we can imagine is close to reality. This is very unlikely to be the case, IMHO. It's fine to believe it is, but that is not science any more. You are entering into religion.
Newton's laws are not known to be "more wrong" than relativity. Relativity matches our observations more closely and is simpler in some respects, but it could be just as wrong. Or more wrong. We have no way of measuring how close to reality our models are. We can only measure how closely they match our observations. As one of the other posters quoted: All models are wrong. Some of the are useful.
Even when you have a model that is completely consistent with observations it does not make it more likely to be correct. Imagine creating some byzantine model with a myriad of exceptions to explain away any discrepancy. Obviously as long as we accept complexity, we can model anything. These models are not usually useful though.
So as much as I understand that you want to ask, "Is this model more likely to be closer to the truth than other models", it isn't a question we can answer in science. We can only answer the questions, "Is this model consistent with our observations?", and "Is this model simpler than another model?"
The article claims that the model is at least as consistent as other models while being simpler in some instances. I think work needs to be done to verify those claims.
We have a current best, or at least competing current bests. These may later be invalidated - that's fine, that's learning new things. It doesn't mean that everything we know now is indistinguishably-wrong as everything we thought we knew in the past. We can predict things now that we couldn't before.
We also have approximations that don't attempt to predict the system, they just produce useful results often enough to be retained - are these indistinguishable from all other attempts at describing reality?
How far do we take it? If Pi is 3, it may actually be true! It'd just invalidate a ridiculous amount of what we think we know. That's equal to quantum mechanics, which appears to have concrete applications.
---
Perfect is the enemy of good.
> The article claims that the model is at least as consistent as other models while being simpler in some instances. I think work needs to be done to verify those claims.
Great!
It's like having a model that says "quack-like sound == duck". It's wrong. We know it's wrong. It may be useful when looking at birds, but it's 100% wrong when your hard drive produces the noise.
Of course it's a spectrum, but some of the things we hang on to and use are known to be false in a large way but still useful.