I think I arrived at the same suspicion independently -- it was when I was trying to understand thermodynamic entropy as an instance of Shannon entropy - where the latter is defined abstractly as a property of probability distributions - which left me wondering about where the thermodynamic probabilities came from. I was wondering whether they were supposed to be subjective probabilities, or derived from ensembles. Then I recalled that entropy was originally defined non-probabilistically as dS = (1/T)δQ. Then I started reading about Boltzmann distributions as a bridge between Shannon's entropy and entropy in the earlier sense (Clausius entropy). I then concluded that instead of thinking about bits and bytes, it was much easier to think about gases and machines doing work, like a 19th century engin-eer building, er, engines.

I am pretty ignorant of this field.