Raw LLM Responses

I love using A.I. art generators but will not use any model trained specific to any artist. I also avoid using artist names (maybe that will change when I know artists that have opted in to allowing their style in the machine). From my understanding diffusion models can actually be trained to forget, it is most likely to result in a broken model but there is a foundation there for machine learning to have the capacity to scrub specific blocks of data in the model, not an easy task with current accessible technologies. I also think that any company that is using an 'opt out' mentality is doing s serious disservice to the A.I. art generation world as it gives all us users a bad name. The biggest asset I currently see with A.I. driven art generation is in the inpainting methodology where I could take one of my photos and mask a specific area and give a prompt for that masked area, example: picture of myself in a yoga posture and I mask an area around the hands and give the prompt 'lightning bolts shooting from center towards outward edges'. It may take numerous iterations to get what I want, and I have spent hundreds of hours on making personalized pieces. There is a legitimate artist side to using these programs but putting in the prompt like 'make a Marilyn Monroe painting in the style of Sam Does Art' or any other artist is definitely not a legitimate artist use of the programs and should be considered a direct copyright infringement, especially if utilized for monetary gain. I see these A.I. tools like a camera or a gun, in of itself it is ineffective unless you have a human using it, and that human will use it with a specific purpose. I believe that A.I. tools could be sufficiently trained utilizing only public domain imagery. The problem that will likely continue beyond any laws and regulations is in private users training personalized models specific to copyrighted information regardless of what the base model is trained on.