I mean it’s literally like, input, convolutional layers do something, output. Of course there is a lot more work to make the output “good” like in the case of chat gpt but realistically the building of even a decent model can be done by anyone that watched a couple of videos online and has some knowledge of computer programming . Figuring out how a model reached a specific conclusion is borderline impossible due to the complexity of all of the hidden layers. If devs could tell you how their model came to that conclusion and what was the models sources/“inspiration” as op may call it, I think at least I would feel a bit better about it. I really hate the “people learn from other people’s work why shouldn’t ai” bullshit equivalency it really is fucking hot garbage. There are artists, musicians, writers that get found out all the time for using someone else’s work and need to give credit and possibly monetary recompense all the time, but when ai literally scrapes copy written sources without permissions it’s suddenly the same as me paying to read a book and writing some derivative garbage cause I’m an idiot.