Minsky isn’t talking about all models, just those used in economics. If you put too much load in the wrong place on a bridge for example, the models will predict that the bridge will fall down and invariably it does. But there the “if” is based on observation where as economic models are based on an assumption. Minsky is saying that the assumption is not only wrong, but is actually causing the bridge to fall down. The movement of particles at a sub atomic level are, as you say, probabilities, by from those observations we can build computers that work, whereas our economy, for the majority of people at least, is not working, its collapse a few years ago was not predicted by the models and the increased austerity and privatisation ever since, also the consequence of modelling, has failed to produce something that works.
Ever more complex revisions of the model are designed to protect the model, not the bridge.
Minsky says that our economic models are built on the assumption that markets, if left alone naturally tend towards equilibrium. Minsky says that such an assumption is not only wrong, but is the problem. He describes a situation where as an economy is stabilised after a recession then we enter a speculation stage where rather than making “safe“ loans where the borrower can cover both the principle and the interest, the banking system issues riskier loans whereby only the interest is paid and the principle is re-financed (interest only mortgages being an every day example)
As confidence in rising asset prices continues we enter the Ponzi stage where more and more speculation happens until the bubble bursts and investors can neither afford the principle nor the interest.
In the example of a pressurised system Minsky would say that Banks are not passive actors or occasional regulators, but are acting as pumps, causing speculation and investment bubbles.
The modelling is causing the bridge to fall down. The modelling might then rebuild the bridge with a few tweaks to redistribute the load, but that produces a misplaced confidence to increase the load until the bridge collapses again, then develop another more complex model, and so on and so on.
The podcast is worth a listen, even if you don’t agree with economists like Steve Keen and the various bankers in it.