I'd like to state hypothesis (sounding like thesis. But please think of it as a start of discussion more than opinion written in stone) that ChatGPT won't produ

I'd like to state hypothesis (sounding like thesis. But please think of it as a start of discussion more than opinion written in stone) that ChatGPT won't produce original work of general physics-science theme because it doesn't have contact with reality. I think ChatGPT is/will be capable of 'understanding' the world/reality in as much as people will write about it in the internet. But by itself it doesn't observe nothing and has no thoughts about the world. If human won't observe something then ChatGPT won't observe it neither. Not my main point but to be more nuanced I think ChatGPT could produce *some* amount of original science in places where people written about some phenomena but weren't able to connect the dots (come up with analogies shedding new light etc.). But ChatGPT won't go out looking for the dots. I think it won't go looking for new dots because of 2 reasons. 1. It can't. It operates on language alone 2. It doesn't ask questions by itself. It only gives answers. It doesn't have the drive to seek knowledge. I'll try to illustrate my intuition: There was a gorilla Koko. She was taught to communicate with humans and could answer questions. But Koko didn't ask questions. She was observing the world but didn't have the concept of unknowns. The sphere of things she didn't know simply wasn't existing for her. I feel like ChatGPT is the same in that regard. 1. I can't find the article. I can be mistaken about Koko, but my illustration still stands. 2. (So many tangents....) Maybe 'curious' model could be built, but ChatGPT would only be a language part of the bigger model. Whole other paradigm would be responsible for curiosity part Back to main point. I have problem with mathematics in the context of ChatGPT coming up with *original* work. I don't know what is the current consensus in philosophy of mathematics about it being discovered or invented. If it's discovered (the fact that abstract mathematics seemingly not describing anything real happen to be used 50 years later as description of some part of reality would suggest that) then my 'not thesis - thesis' stands. It won't produce new original mathematics. If it's invented then it won't need contact with reality and it (theoretically?) could *creatively/originally* advance mathematics (once it already exist). Question. Mathematics originates in axioms. Maybe that's why it fits the world. Because people observing the world chose the axioms that made sense to them. So when it's advacing to it's logical conclusion it still happen to fit the reality. If mathematics wasn't formalized already would ChatGPT ever invent mathematics? Would it come with any kind of axioms?