I've never met anyone capable of communicating with well-defined terms. Or giving definitions that actually match the real-word usage of the term. And all definitions are ultimately circular, because the number of words is finite. In any chain of definitions, you will eventually have to use a term you were trying to define.
What you call the scientific method is a philosophical construct that has little to do with actual science. And philosophers disagree on whether it's a good ideal for science. Given that it's neither a good description of science nor a universal ideal for science, I wouldn't focus too much on it when discussing science.
> And all definitions are ultimately circular, because the number of words is finite.
I can’t help thinking I’m talking to an LLM or a troll.
If you use a complex term that needs definition in a casual discussion, it’s most likely none of the words you use in the definition would themselves require definitions—and if this was to happen repeatedly, the conversation would halt long before we would be running out of words. It’s enough to avoid circularity within a couple of levels in good faith.
Anyway, I’m not sure whether we disagree or not or what exactly we are arguing about. My point is “ML making predictions is not a threat to us getting at underlying principles, because natural science (scientific method, predicting things) in general does lead us to any provable facts about those principles, and because ML would make predictions within an incorrect/incomplete model that we gave it.” In that, by “underlying principles” I mean some statements about objective reality. If we are on the same page here, we can continue discussion, otherwise let’s not.
It's an analogy. All communication is ultimately circular, and we can never be sure that we understand the terms the same way as the other party. Still, people often seem to be able to communicate.
Similarly, scientific method cannot discover the underlying principles or explain the nature. It can only rule out principles and explanations. Regardless, science seems to come up with principles and explanations all the time.
And that's because scientific method is not science. It's a theoretical model for a subset of (often ritualized) activities within science. Actual science is more than that. It can do things scientific method cannot, because it's less focused on philosophical ideas such as absolute truth or provable facts.
In my experience, scientific method is like a picture of an elephant drawn from a written description. Drawn by someone who has never seen the animal or a picture of it, and who has no idea what kind of an animal it is. There are some recognizable elements, but it definitely does not look like the real thing.
What you call the scientific method is a philosophical construct that has little to do with actual science. And philosophers disagree on whether it's a good ideal for science. Given that it's neither a good description of science nor a universal ideal for science, I wouldn't focus too much on it when discussing science.