The Validity of AI-Generated Visual Art

A friend of mine has given me several really good ideas that I managed to turn into excellent photographs. She has an exceptional visual imagination; what she doesn’t have is the skill set to allow her to realize her ideas in the real world. And while she has become a very good photographer, she again lacks the technology and skill set to create what one would call fine art photography.

Recently, wanting to realize an idea that she had had to create a special image for a friend, she decided to turn to AI to see if that would help. After all, she was quite good at explaining her visual ideas in words and thought that perhaps AI could help her produce the results she sought. Her first try was not a success, although there were enough pieces that were right, so as to make her hopeful. She revised her prompt and got better results. At one point, she included a reference image in her prompt. The results were even better. After several more attempts, she go results very close to the image in her head. Deciding that this was as good as it was going to get and being limited by having a free account, she saved that image, and performed a face swap on it using the tools that were available to her, resulting in an image that was very close to the image in her head. It was good enough for the purpose for which she intended it.

Somewhat concerned about the ethics of the whole procedure, she then began to do visual searches for elements of the finished image to see if she could discover the sources for the image she had created. After what she considered exhaustive searches for elements of the image, she found no discernable sources. Satisfied that the image was as “original” as possible, she accepted it and sent it to the intended recipient.

We all know that AI “learns” by taking bits of information, in this case visual information, from the internet and any other sources at its disposal. In that regard, anything that it creates is derivative, and if the product is derivative, there are ethical considerations. However, even the most original human-created work, whether visual or otherwise, is somewhat derivative: creators build their work out of pieces of works they have encountered. The trick is putting them together in a way that no single source is discernable and the manner of assembly is unique and original. Unless this occurs, the work is considered plagiarized to some extent. For work that is not plagiarized, careful analysis can discover “influences,” that is derivative sources, but none of which are used to the point of “copying.” Indeed, while some influences are conscious, many of the influences on a work of art may unconscious on the part of the artist. They are simply part of the artist’s background—work that has been encountered and assimilated and therefor useful in developing original work.

So the question becomes how is AI, used ethically, different from any other artistic tool? There are a plenty of images that can be found with a quick internet search that appear to be totally original that have been created with AI. It just requires a different skill: the ability to write a prompt that pulls together disparate elements into an original visual work. Thus, it appears, that AI visual work can be as valid as work created with any other set of tools, plus a good visual imagination.

Date: Sunday, 30. June 2024 22:08
Trackback: Trackback-URL Category: Originality

Feed for the post RSS 2.0 Comment this post

Submit comment