Raw LLM Responses

​@romanpyatibratov4361 That kind of training would likely take years just to get one image. Maybe even longer. It would be stupid, inefficient, and frankly infeasible. "Technically possible" doesn't mean anything. And it doesn't even matter anyway since all existing models _did_ use tons of training data, so what you're saying is just a hypothetical. Like I said in my comment, I could not care less if the output might be able to be compared to human output in how transformative it may look. Meaning, it doesn't matter if the specific output didn't exist before. The whole entire issue is the usage of stolen training data. Yes, we humans need "training data" to be able to actually replicate things, but as I said, we have conscious, thinking, experiencing minds. We can look at the source data, think about it, and make it entirely different. We could see a human person, imagine they had long triangular ears, and say "I've now invented elves". A generative A.I. model would not be able to do that. A human person would need to feed it images of elves, a made up concept by us, to be able to reproduce it. Or, of course, go your route of endlessly telling it do iterate and iterate to get that same output. But as I said, that would be so time consuming it would be practically impossible. A generative A.I. model doesn't have outside experiences. It doesn't have emotions. It doesn't think. It can't see. It can't hear. It can't taste. It can't smell, It can't feel touch. It's a program. A complex set of equations that tries to accurately predict the most likely output given the prompt and all its training data. It gets an objective, then performs the objective. Training data affects its parameters to be able to reproduce images, and that's all it is. A.I. with their neural networks work very similar to our brains, in a much more simplified manner, but the fact of the matter is that _we are not just our brains._ We are everything else too. Until an A.I. gets a physical body, gets multiple constant sources of stimuli (seeing, hearing, touch), and is constantly running thought processes (not just on standby until prompted), I won't think of data training as "taking inspiration". Just theft. (in cases where source data was taken without consent) So, kindly, remove your Clippy pfp. You're a disgrace to what that movement stands for with what you just spouted. It's literally a movement against companies taking people's data without their consent for their own gain, which includes training A.I.