It's basically different terminology. His point is valid.

A model isn't something you assign probability to. It's something you use to come up with a set of prior probabilities. The model he used assumed independence. It didn't actually assign zero probability to any result. It doesn't assign a probability, zero or otherwise, to the machine being broken, because that's not something that's considered. It also doesn't assign a probability to whether or not its raining.