As a working artist in the VFX/Game Industry; I was sharing with a friend a piece of work that had been produced by an AI from a Youtube video and he as a working modeler became concerned seeing huge part and swathes of his work being automated; specifically as an artist in smaller mobile/independant studio.
The video in question; he's done a lot of textures works on mobile games.https://www.youtube.com/watch?v=1VHCIKMrWHk
However I did remark to him that the finished result that the AI Gave was akin to a Baked Out image from which it became very difficult to make minute detail and change since you had to deal with a finished image you could only paint-over on and not a file with sets of layers where you fine touch and tweak parts of it layer by layer, group by group.
Edit; Decided to change thread's overal direction,
I think this sort of thing is going to hit outsource providers and those working on shovelware first but it'll spread out as it gets more controllable and if you're not able to do more than the robot you're going to find yourself getting replaced.
Alex_J said: Does this work? My friend asked
This one here's getting a lot of hype too but behind all those flashy generated assets and decors, there's a lock of control and fine tuning.
The biggest questio now is how will all those AI tools give back that kind of control/trackback and optimization in the future?
There's a lot more money in providing tools for the masses than there is in providing tools for the creatives (there's a lot more masses than creatives) As a result, the current pile of hype-driven generative ML apps are aimed at providing a finished image rather than a useful tool and are utterly useless outside of generating some fun ideas for a concept artist to paint over.
When the hype and rapid-return investment panic dies down we'll be left with a load of interesting tech that can be redirected towards something actually useful and we'll start to see tools that legitimately extend the capabilities of artists.
ML is extremely powerful in terms of replicating and predicting the behaviour of systems - this allows for all sorts of very abstract possibilities but if you want a couple of simple examples that aren't threatening to the artist...
a: its a perfect slot-in upgrade to the (largely noise based) procedural systems we're all using today
b: it can do simulation (cloth/fluid/soft-body dynamics) much, much faster than running the sim traditionally
Personally I'll never pay for anything AI related. If they stole from us, it is our tech by right. Just like tech companies dealing in user data. If they take our data without our consent, we get to use their services for free. Lucky for us that we're the most suitable people to use it.
your game will be great man, you won't fail if you make the game you want to play.