I think AI replacing artists is going to be a good thing in the long run.
Right now, if you want custom artwork made without the use of AI you have to either criminally underpay an artist on Fiverr or pay an artist hundreds of dollars to make one piece which you may not like.
I think the cat is out of the bag on AI generated art pieces and there’s no way to put it back in, and so future artists will use a combination of their actual art skills and their ability to work with an AI system to create entirely new and currently nearly impossible art pieces, and there is an entire field of unexplored possibilities waiting to be tapped by Future artists.
AI can’t make new art.
All it can do is repurpose pieces of art other people have made, just like Auto-Tune can make you sing on pitch but it can’t make you a good singer.
Like with many technologies, it creates as much as it destroys. It is a tool, tools change how things are used, some are adapted, some fade to obscurity.
I think there are merits to LLM’s, like for games and obviously translation. I also think for games there will be a script director who would work with the games LLM to make it work in the games universe (rather than spewing random things the player said in a playthrough).
I also think there are merits to AI generative tools, however I think they will be used in ways that differ minimally from something like procreate. Given that things like infill and nodes have developed so far as they have, it’s really just a few steps behind the “polish” of an adobe product. And mostly open source (depending on what you use).
Finally, I think something people tend to forget is that commissions are for people able to pay. If someone is creating free art because cost is an inhibiting factor then were they really going to be paying an artist for a commission?
That said, there are still downsides, and a vast number in our current state. Chances are high that copyright laws will go to protect corporations that use AI but prevent people from using local models. A far better approach would be for independent artists to be able to create, license, and sell their own checkpoint models, but that will never happen for so many reasons, primarily artists having shunned AI very quickly.
Another downside is that it’s just so damn resources expensive. Running AI on local hardware had gotten more resource efficient relatively quickly, but it’s still so much higher than it needs to be still. We can get the exact same results, better and faster even, with far lower power consumption by just using an analog computer to sift through the model and translate that analog data to digital data. So even if someone is just running it on a laptop with stable diffusion low beam settings it’s still quite a bit more intensive than it needs to be. Frankly, tensor cores and CPU’s are a brute force method.
Still, at the end of the day AI is a tool with a relatively low skill needed to get adequate results, with a very high skill ceiling for making something truly good with it. There is nothing wrong with a 2 job 60+ hour work week Sally wanting to use midjourney because it’s the only art she can currently make. She deserves that.
At the same time, if it’s true that Disney used AI for the Secret Invasion intro - what the fuck? If that’s replacing an actual artist, or an actual artist was trying to use Deforum and had that as finalized work that is a much different situation. As with all things, set, setting, and moderation.