Raw LLM Responses

The AI's existence that is fed on this data is not in itself unethical at all. This is a tool after all. What people do with it is unethical. What needs to stop is people making money by creating art with your name on the prompt. You're right, that's borderline harassment. However, it's not nor should it be illegal to scrape completely open information off the internet for the purpose of training a machine learning model as the image's data are in no way transferred inside the AI itself. What is transferred is the connections between word descriptions and the features of the image without the features themselves being saved anywhere. You can make infinitely more images with the AI than you used to train it. Yes, it is exactly like a person seeing an image and then afterwards trying to be inspired by it, but way less complex because of the incredibly more complex human nature as you mentioned of course. Same concept exactly though, just scaled down. The reason the music industry has less problems about that is because there are way less ways to get copyrighted content in that industry for free (YouTube runs add per view/listen so it's not considered free in this context since the artist still gets paid royalties) and is also protected by the policies of the websites that upload said content which might include web scraping limitations legally.