"Mistakes" and "uniqueness" are human concepts that are unfairly applied to AI art. What exactly is a "mistake"? Is it when it draws extra limbs or fingers? Is it when it draws an uninspired background? How is it supposed to know what the mistake is when there is no intended goal? It's not like anybody directly asks the AI not to do these things nor give any incentive to not do them. Even in the case of games AI it is geared towards AI making the best moves, never the most "human" moves, nor having a particular preference beyond "easy/medium/hard" which limits the creativity it can output. To the AI it outputs what you put in and it hasn't made a mistake.
"Uniqueness" is another standard unfairly applied to the AI. What's unique? Is it Fern with purple hair and a white dress that casts magic? Or is it patchouli knowledge with purple hair in a white dress that casts magic? How exactly do you define what is a unique character to a person who has never seen these two characters? If the AI generated these two characters randomly would this person who has never seen them before know what is unique or what would sell best? If you went to the best human artist and told them to draw a purple hair white dressed mage who casts magic would they be able to make a unique design or would their clout as the best artist make that character popular?
I always bring up ZUN in these art examples because he's not only the most unique and influential artist in our current era but he's also someone who's art is considered "bad". Yet despite being "bad" his designs make up the majority of the fan art posted on Danbooru which is one place the AI is trained from. And despite being a very unique artist he had based his designs on touhou 6 on a manga he read. So we want the AI to be unique and successful like ZUN, yet it can't make the same mistakes in art as ZUN nor take inspiration from already existing works fed into it? It's an impossible task and part of the reason why much AI art seems "soulless" because the human user doesn't know what they want.
>so ai would, at best, be incredibly stagnant due to the fact it's unable to go beyond human creativity, or at least that's the case due to how people have built art ai
Another impossible standard. What's beyond human creativity? I wouldn't know I haven't seen it so how would I know it went beyond human creativity?
The problem with people's opinion on AI is that people are over estimating what it should be doing rather than treating it as a tool to assist or play around with.
In my best application of AI I've been able to have it make images that I can't make nor could the AI make itself through prompting. For example I made an original mech using cropped images from Google search and in another case I had the AI generate a Palworld character using nothing but a low resolution reference image. The true power of AI exists in people that can understand its limitations and can maximize its output, and so far people are not creative enough to understand that just yet. #zemAI
{{{White background}}}, no humans, s-638094618 (4) (1).png
{{{{{realistic, photorealistic}}}}}, black background, no humans, standing, robo s-2196106946 (2).png
_{{Realistic, photorealistic}}, hangar, science fiction, s-1026248445.png
game cg, 1girl, outdoors, cloud, colored_skin, covered_eyes, flat_chest, hair_ov s-839404309.png
