I feel like we're getting a good picture of what generative art (at least this kind of Midjourney nightmare) undermines when it comes to humans making a living. The stylistic mimicry is amazing, but the faux-surrealist incoherence is obvious and difficult to avoid even with clever, iterated prompts. I don't think our artistic eminences have much to fear from stochastic infringement because the AI has no understanding of what it's painting, only an abstruse technical mechanism for generating superficial and uncanny derivatives. Videos such as this, though, are surely not disclosing the full prompts used. I bet a lot involve named artists and it's funny that this line comes with an image so evocative of Simon Stålenhag's abandoned vehicles and long lenses:
If AI art is more likely to clear the orbits of established talents than to convincingly automate them, then, it limits the market for inexperienced artists who might one day compete for well-paid work. Artists who might once have established a professional profile for themselves by sinking 1000 hours of hard-attained skill into a project such as "I painted surreal art for every line of Styx's Renegade," instead of the 10 hours of prompt experimentation (using mostly-proprietary, VC-funded technology) represented here. Consider also the sort of cheap dystopian shitposting I no longer need Photoshop to crank out.
AI art is full of tells and the tells signify automation. What it can't do will become fashionable: precise work where all the details emerge from a structured and coherent model of the world or the mind. Good news for skilled drafters!
Previously:
