gizmodo.com: One of the main defenses used by those who are bullish on AI art generators is that although the models are trained on existing images, everything they create is new. AI evangelists often compare these systems to real life artists. Creative people are inspired by all those who came before them, so why can’t AI be similarly evocative of previous work?
4 comments:
It is scary in what AI art is capable of doing and sometimes, what it is not capable of doing. From the beginning, even though I knew that these machines were using pictures and art forms that they memorized to generate “new” ones, it is still scary in how fast that it can create a series of artwork. With this article, this idea of copying existing artwork was ensured and further proven true. This research is definitely comforting and provides artists an explanation in why these machines is capable of doing these works quickly. But it also brought the question of copyright and who owns the image. If it is the creator of the AI generating machine, it seems unreasonable since their work originated from other artists pieces. If it is the original artist, then what about the parts of the artwork where it was generated and inspired by other artists.
I’m not surprised that this is the case, I think that the power of AI is something unlike anything else, and I firmly believe that AI can do some pretty awesome things. But I’m not surprised at all that it can just copy existing images but reproduce them with marginal differences. I can see how this isn’t the worst problem in the world as the article mentioned that software could be developed to flag these types of duplications; but I wonder how long this would take as the duplications are already occurring within certain software’s. I also wonder the affect this could have on our industry – we see so many articles about AI in shows and their potential, but what happens if any of that substance put into shows was a duplicate of somebody else’s work? Is this a possibility? Would that above flagging software have to become popular before AI majorly integrates into the industry?
This was an interesting read - I’d never really thought about how AI generations can just end up pulling one image to create their “new” one. I’ve thought that AI art does the copy and mix implementation in order to create a supposedly unique new image, and this finding does partially correlate with that train of thought. AI cannot physically create a new image - and I’ve seen plenty of artists on the internet showing how close an actual art piece they made was to an AI generated art with a prompt. There are certainly a lot of risks with AI scraping information off the internet to create new images and this article presents a number of them. I’d be interested to see if there is a way to prevent this duplication or close duplication from occurring at all but I’m sure it’s very difficult. AI in the end, is just creating from work it has found, and this article highlights that.
I have heard of the discussion of AI generated artwork, but I did not realize how much the programming allows for replication.
Reproducing art is not a new issue, but the scale in which AI is able to steal work is very concerning. My cousin is an artist and she has had on multiple occasions had people reproducing and selling her art or creative ideas, and she can only do so much to discover and disway them.
Many people on Instagram or Tiktok have seen the infamous ‘anime filter’, that turns a person in a photograph into an anime character in a matter of seconds. It makes sense why people are having such a good time with this filter: users trying to trick the system into giving them huge breasts or seeing how it would react to a photo of a pet. I am worried about the implications of AI generating photographs of people doing things they have not actually done.
Post a Comment