The presence of mistakes in art are part of what distinguishes human made from AI made art.
Human mistakes can be indicative of the effort exerted making the art; human mistakes can also be a conscious inclusion by the artist. Mistakes can be an artistic technique, such as splattering of paint by Pollock, in that, the actual application is partially random.
Another example of a conscious mistake in art would be paintings by Klee that include a grid. His grids are always off kilter, neither even nor symmetrical. What appears to be simple, perhaps childlike, is anything but to AI. The computer would have to calculate each angle separately, which of course it can do, but that is antithetical to how an equation wishes to function. To create art like Klee, the AI must first study Klee, then mimic him.
Human mistakes may be intended or unintended, either way, they are part of the artwork. Some are not immediately apparent to the audience, some of whom will never apprehend the idea of the error being included. A mistake’s inclusion is a challenge to us to consider the creation of the art; the art is partially a statement that the artist chose to include the mistake.
AI mistakes are a result of incomplete data and lack of oversight. AI cannot make a mistake unless prompted. Adding noise, data or signal corruption, competing logic loops, other forms of generated interference are not mistakes. Mistakes cannot be programmed. Simply put, AI art with mistakes is just that — something posing as artwork with inconsistencies, made as a result of bad image management.
There is only one way to make a mistake that works and that is to be human. The simplest definition of this form of human mistake: a beneficial result that is not always the intent of the effort. Which, can in turn, show the art is human made, and that adds rather than detracts from it. This distinction can be shared by both the worst and best human art, because humanity is no guarantee of mastery. But it will always show, whether good or bad, it was imagined and realized, which AI cannot do, only mimic.
We look to computers for perfection. We allow humans errors. Computers, until they possess full consciousness, are only machines and operate as such. AI, at the moment, is only a more complex program, not true intelligence. When AI can create art that contains inconsistencies, small errors that might provoke inspiration to incorporate those errors, or even more advanced, a purposeful inclusion of ‘mistakes’ which could also be termed ‘controlled happenstance’ for no other purpose other than the joy of the making of the art, then AI will have accomplished a step toward full being.
We need to learn to distinguish between human made and AI in art. Mistakes can be a vital clue — the inconvenient data that does not belong. Plus, the unmistakable sheen of artificiality all AI currently generates, is to my eye, unmistakeable. Superficial, plastic cosmetic simplicity, no matter what the artform, AI cannot generate the excitement of making the human artist experiences and expresses.