random thoughts about AI art
a problem i feel like we've got already with discussions about AI art are people who are ostensibly on the right side of the argument, but are so excited to fight the good fight that they end up being completely intellectually incurious about what the actual problems are.
"real art is made my humans, not computers" is a argument that throws a lot of helpful, efficient tools for artists out the window. this argument as applied to this situation cannot be used without procedural generation as a whole being raised. all of a sudden using blender's geometry nodes is now potentially invalidating. speedtree is a sinful tool for all game developers.
"it's always looks wrong and gross and creepy" is an argument against virtually all outsider art. wrong and gross and creepy are very useful tools in the artists' tool box. the uncanny is the striking, or the means by which we invite the audience in with comfort only to have them discover something shocking with their own examination. of course, i'm making the argument for its intentional use, but to use this against someone who tries to create beauty and lands on ugliness is an argument against every artist who is still learning.
"it can't even make hands" is a neighbor to the last argument, basically the idea that these systems will always be imperfect in their execution, and will always have telltale signs of what they are. that's simply not true. as long as they're continuing to be used and mastered by the people using them they will improve. they'll become more and more sophisticated and better adapt to modern sensibilities. if you're relying on "real artists will always be more technically proficient than computers" you're volunteering your heroes to fight larger and larger dragons every year.
the central issue (for me at least, your milage may vary, i am not a doctor, this is not legal advice, etc) is that every AI art generation algorithm i'm aware of was trained at least partially on stolen art. this was kind of always a necessity. the volume of data required to start getting useful results was always going to make assembling thousand and thousands of contemporary artworks legally financially infeasible. the engine needed fuel, and it's inventors were reckless about where they got it. now the influence of all that art is caught in the ephemera of these models in a way that it's simply impossible to remove. they will always be the result of this unethical act, and that is the driving argument against them.
this isn't NEVER brought up, obviously. i hear it most often from actual artists who are constantly walking a tightrope between making their art as accessible to the public as possible and trying to preserve it's value and retain any sense of ownership over their work. the general population, people who's hearts seem to be in the right place but lack stronger familiarity with the issue, have developed a habit of jumpin into arguments passionately arguing points like "a computer can NEVER make REAL art," "it ALWAYS looks GROSS and STERILE and CREEPY" or "it can't even make HANDS!!" these are surface-level objections, so reductionist as to not really be very helpful. they're the art-tech equivalent of "Trump is a bad president, he's so OLD and FAT and UGLY."
