Raw LLM Responses

For me human art vs ai art is analogous to terrain modelled by a human and terrain generated, for example, in Minecraft. Sure, Minecraft terrain could be intricate and arguably beatiful, but by itself it doesn't have much value since knowing the seed, the coordinates and the version of Minecraft it can be perfectly re-made as many times as needed. It's not designed by a human, it's found by a human by feeding random numbers into the terrain generation algorithm, which uses this number to decide where each tree should be placed, where should it put a village and how should a mountain look like. AI art is similar - it isn't made by a human, the entire creative process is carried out by the machine and therefore a human doesn't create it, a human finds it by feeding the algorithm with a prompt and a random seed number, making the artwork itself less valuable in a way, since it lacks both the human effort and human control over it. And what is also interesting about AI art is that it's fundamentally limited in a way - the process of making AI art basically consists of two steps: generating lots of pictures and cherry-picking ones that look great. So AI is limited both by the properties of a generator and by the rating of a human. And a usable AI image generator is not able to create any picture, it will be fundamentally limited by the specifics of ist algorithm. It is possible to create an algorithm that will be able to create literally anything, the website "The Library of Babel" did just that with text and there aren't any technical limitations that forbid making something like that for pictures. The problem is that since a library of babel-style algorithm creates literally all possible combinations of data and prasents them in pseudo-random order, finding a good picture that came out of something like this would require scrolling through enormous numbers of pictures (I guess literally thousands of trillions if not more) to find anything good-looking, most pictures will just look like random noise, such algorithm is practically possible, but is also practically useless. So when designing a practically usable image generator, one should always balance creativity and stability and modern technologies aren't that good at making both of them good enough