>artists will just pivot to work with the new tech

I think the point is not so much that it will completely replace artists, even with significant improvements you'd still need an actual artist to come and touch things up and make sure the output is desirable in the first place, but rather if you have 10 3D artists and this new tech makes them twice as productive you can hire half the number of staff.

It may not completely replace artists in anything other than extreme situations, like indie studios that don't mind if the models are a bit iffy or they have to do some work themselves, but it doesn't need to replace you entirely to devalue your work.

It's the same with most new technology. Automation lets you do the same amount of work with less staff, and typically with less skilled staff at that.

Would be nice to get some solid retopo ai models. But I think 3D will take a lot of necessary progress to kill off. That whole realm is such a behemoth with all kinds of intricacies and requirements. It’s definitely not been primed to allow for workflows to be as vague as “input thing, output result”.

That said, I don’t think career art is safe. One of my many unfortunate realizations is that career art is nothing all that different from a regular job, it just exploits creative technical ability. Unless you’re the 1%, start your own business, something like that. Definitely don’t think jobs like basic graphic design, modeling, rigging, roto, are particularly invulnerable, but they’re currently still needed. But if the tech gets there, I get the sense that industries will be eager to have the same work done for a fraction of the price, even if there are hurdles along the way. The good part is: perhaps art can get back to its roots and actually be about creative expression. Bad part is that it may become far less sustainable as a source of income. I would personally love to see people treading CG more like a traditional art form, but it absolutely does not pave the path to getting a job. Fast, clean, done, again is the mantra for jobs. Predictable, efficient, the type of thing that AI can improve upon.

It’s also no easy thing to just slap in a description and get consistent results across all your inputs. As you say figuring out how to get the AI to draw what you are asking it to is a wildly complex task that takes ages to practice and refine for EACH input. Weeks and months in some advanced cases.

That is an artistry. A modern technical artistry but art nonetheless.

This is disgusting. AI is supposed to be this amazing thing but all I ever see it used for is to kill culture. As you just gave an example of, this is set to destroy the jobs starting artists need in order to grow. Artbreeder says on it's home page you can "craft" AI art like never before. That's a dangerous statement. It needs to be absolutely clear, if you are utilizing an AI in any artistic endeavor, your not crafting shit.

People will still create when there is no financial reward — although much less. There still will be plenty of data to train on.

It will certainly limit the creative expression of the human race, but the profit machines that drive this innovation don't care about that. Human expression doesn't drive revenue.

You need to allow yourself to dream bigger. At some point we will create models able to create new styles. Also it's a potential job for artists to be employed just to create things to teach AI on them.

Nope. You can feed AI on a hand-picked database of the best of the best of generated images. I'll give an example: generate one hundred thousand images of hands, have one hundred people sort through all of these photos, doing one hundred images a day. Delete any hands that are outright anatomically incorrect. Then, feed this back into the AI until it learns to generate better hands. Repeat this process for anything.

In fact, this is already happening with Midjourney. They track which photos their users select, then they feed those photos right back into the model. Coincidentally, Midjourney v4 is currently regarded as the best image generation model available at the moment. It's considered to be far ahead of the competition at the moment. No points for guessing why.

I reprocess and sample past AI work for consistent new AI generations already.

I graduated with a computer science degree in 85, and at the time prior were working on auto code generation. The were articles saying programmers would become unnecessary. Nearly 40 years later and auto code has gotten much better but we need more programmers than ever.

I'm guessing you won't be replaced anytime soon, though your tools might change.

Look at what the actual output of that is, it's practically unusable. I've done work in this field, and while this is genuinely impressive and a step up from previous work. It doesn't really have any practical use beyond rapid exploration. Arguably the most impressive thing about this is the natural language processing to 3d synthesization, and the speed at which it does it. How this would be used is to quickly synthesize rough ideas, modify them, and then send them to outsourcing alongside concept art guidelines for final asset production.

Production of 3d assets these days is already largely outsourced, but art direction and dressing is still going to be needed. What will be truly revolutionary if an one could derive context, style and intent behind an existing scene or concept piece, and just feed you models or 3d elements that would best fit within the environment or character. Not even synthesization, just asset suggestion is already a gargantuan task. Promethean still largely works off natural language and meta data, but not from one shot image or geometric data, for now. But you still need a person to properly direct towards a desirable output.

We're still miles off from replacing actual artists.

Retraining to do something that will be equally displaced by these developments? ;-)

Don't be terrified, be open to whatever and however it will affect the craft, incorporate it or don't (there is space for both).

Change is constant and when there's nothing you can do about it, there's no need to worry. If there is something you can do about (or despite) it, there is no need to worry either.