AI is going to be our biggest game changer
This discussion has been closed.
Daz 3D is part of 
Connect
DAZ Productions, Inc.
224 S 200 W, Suite #250
Salt Lake City, UT 84101
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2024 Daz Productions Inc. All Rights Reserved.
Comments
Cited the wrong post, meant "Since so many here seem to understand A.I infinately more than I can a perhaps realistic render of a Daz figure be processed through A.I. to make it more realistic? I'm not even certain any of them take direct image inputs. the course of this thread made me curious."
Well, in a million ways (say: two, three at least). The generative kind of systems have inpainting abilities, so you could more believably put images into other images. Don't ask me how exactly to train a system to achieve what. For a part one doesn't really know (precisely), still the language model part makes it accessible via natural language (somehow). There is not a minion, who thinks or considers and reconsiders, it's still "statistics" for most, in a way a modernist encyclopedic view onto lots of training data. If there are no 2-legged entities in the training data, it will be hard to create any, similar for anything else on any level of detail. There is some synthesis and abstraction (statistically!), but it's not comparable to real-time recurrent, reconsnidering and reiterating real-brains. You CAN train such models on DAZ models, and i assume that it'll happen in one or another way (generically, large-scale), at some point. Training of large data sets is expensive, though. And you would count in some rendering time, from multiple angles, and then morphs etc... the complexity is too much even for a hype, but with some smart [to be elaborated/invented] concept, you could do "A LOT" in that direction, if "you" are DAZ. If you have a lot of library, similarly, if you have the skills and time (...). Nifty less expensive? Maybe possible, for specific tasks, maybe even more feasible with science progressing...
@WendyLuvsCatz "I also think people using premade assets complaining about 3D art using their images to train are in glasshouses throwing stones " - No it's in no way the same. You PAY for a license to use stuff. Stuff is being created for YOU to pay for a license to use it. Further, imagine i put tons of money into DAZ assets, wouldn't i want it to stay at least alive enough for me to use them forever? And if i die i might not even give it away. Or can i?
@Gogger "While not everyone feels the same, for these people A.I. art feels "lonely". " - Lonely is the essence. AI art isn't necessarily :p. Hmm, what was i aiming for...
+1
@WendyLuvsCatz
Agree completely.
This AI tech is about being integrated into everything, even the mighty Blender.
any creative software company that tries to ignore AI, does so at its own peril
https://github.com/gd3kr/BlenderGPT?fbclid=IwAR1Qqx4XhlWnZE5Lu84n98iECgMBd90IVBwyM_ro4H9tmGzrJP2AgyDv5YQ
You're forgetting that the 3D content is LICENSED. Whether paid or by the graciousness of the creator to provide it for free, they are AWARE and have given CONSENT to their use.
A.I. datasets used by Stable Diffusion, Midjourney, and the like were NOT. These were only for RESEARCH and EDUCATION. They were not intended for commercial use, but these unscrupulous companies took them anyway to FEED their machine learning because it would have been too costly both time and money to acquire permissions for the data they needed for their primitive A.I. (yes, it's primitive... it cannot function using the denoising process without the hundreds of millions of datasets they conveniently mooched).
In any case, your point is moot. 3D models are now targeted with SketchFab being a recent rider of the A.I. bandwagon and the artists are angry because their works are being used in this predatory practice and they are unable to opt out.
https://twitter.com/xrossfader/status/1639372724480143361
You are incorrect. Artists that licensed for the Adobe stock state they did not consent for this use. Moreover, Adobe is preventing them from opting out, which is a legal issue and lawsuits are being prepared by the artists affected.
https://twitter.com/eballai/status/1638603144455135238
https://twitter.com/MarMai1/status/1639121762406236161
https://twitter.com/kat_loveland/status/1638707061889564673
yeah that's a shame, I was hoping Adobe was big enough with licensed content of their own to produce a truly ethical training model
they claimed it was
I guess the lawyers are going to make bank once again, maybe affected artists too if lucky
What about Nvidia Picasso?
Hope they have AI content trained on images with the proper license.
NVIDIA Reveals REVOLUTIONARY AI - Better Than GPT-4 And Midjourney V5
I want an ethically trained AI too, we all can agree on that
what we cannot agree on is people such as myself using what we have for now even for self exploratory educational purposes without profiting from it
It's like, yes this wonderful technology exists but don't you dare touch it and learn from it and share your findings because bad, naughty, thief
I feel like a black magic wizard using dark powers TBH and it's not getting the result the haters want
anymore than telling me not to eat the double fudge chocolate cake would
I have a sneaking feeling that even if we get a fully CC0 trained model they are still going to say because it's not human made art and taking the job away they didn't get anyway it's evil
Test of Adobe Firefly on YouTube:
The Prompt Wizard doing dark arts AI majiks on their iPad
With AI; unreal's FAB; photogrammetry; AI + posing tools; AI + scene building; AI + add-ons it makes me reconsider what to buy now / what to mix together for sale, games...
Pretty much in a year, we will see how things will start to integrate into unreal, blender, unity(?), and Daz(?).
Youtube is full of videos of AI add-ons that generate scenes, textures, posing, for that we would need good indexing of the store, assets, and tags.
Probably some PA-s will try to do 2D pictures to pose AI too in DAZ.
Good question is the scene building as tags etc... is not a strong part of DAZ and it would need a lot of back-office work.
Is it ok to download content from warez sites to "try it out"? That, or other variations, are common attempts to pass as legit on such sites or by such users.
until a court of law says using AI is illegal this is a purely moral argument like sex outside of marriage, eating pork or going topless on the beach
if and when it is made illegal then I will cease and desist
IMO whilst rendering in Daz or Poser we are on the same level, asset creation wise, with AI generators. The only functional difference here is for users who do build and texture more than, let's say, 50% of their assets. The legal argument here could be if it's provable that the AI in question was trained with illegal assets. If it's not, then a fee paid to the AI service is very much like paying for 3d assets.
The truth is, if you want original art then you have to either build the assets yourself or draw/paint from scratch. Build everything you need the hard way.
Except for maybe contemporary and historical story settings, but most of the Daz contemporary assets are glamour and actual historical props and costumes are few.
Licensing assets still is different. A cloud machine is one company profiting over everybody else - everybody, including those who gave training data. It's currently not feasible to pay anyone for it, because you need so much of training data. That's also why most players with actual commercial interest screw over the people who created the training data they're using, either directly, or with opt-outs and further tricks (at least it very much looks like that). This is just "natural" for the tech-people, because the technique, in terms of science and CPU/GPU/accelerators are ready now, but you don't have the curated data sets built up for years or even decades. So NOW+FAST+CHEAP is the essence for the big scaling things.
In contrast to that, we both pay DAZ (tools) and the creators directly by buying assets, and we further keep the ability to use those at our own peril. This means some freedom, which might not be the case with a commercial generator cloud service, conceptually. Of course there already are and will be many more ai-based tools, that work for enhancing whatever you do, with your assets, renderings, stick figures, whatever.
Maybe the markets are over-saturated and over-heated, and maybe there exist too many PAs :). Still if all people go for the cloud service (or too many), the tools and asset creation might suffer, in terms of the number of people who can make a living with such (or partial), the price tags, and in the worst case: abilities (+tools) get lost.
Concerning original art ~ typical would be to draw it youself (cough), or get it done for money (cough again) - you could now argue that DAZ is the run-hot version of a democratized state of things, or it's just the benefit of the internet, so the technology allows you to make art cheaper for more people, instead of charging 1k for the random toon character for some commercial. The generators in a way democratize further, in the "social media" sense, provided they make it even more affordable to create interesting images with even less skills and time at stake. However monetization-wise, this certainly is a drain with one pool to fill, and there is no PA in the pool, hardly any toolmakers, and in the end you can't be sure, if you yourself will be able to use the cloud-tool the next day. So freedom exchanged for time (also could count as freedom) maybe money (also could...) in exchange for loss of control (certainly less freedom). The equation may yield different smilies for different use cases or people...
(No PAs in the pool - for stock photos you might try to go for assignability, but i haven't heard of a way to do it with the generators, yet. It doesn't even seem to be feasible technically, thus the money anybody gets likely will be very much marginal, e.g. if they share a couple of cents for all images that are tagged similarly. Tags vs prompt, but then quantification? Hard to imagine...)
I said "asset creation wise", not legalities because that has to be resolved. Adobe, Nvidia, very likely DeviantArt and ArtStation are going to at least try offering paid services if they can solve legalities of it (arguably, given the emergency evacuations from DA, AS and now Adobe there's going to be harder to train the AI, but they will come up with something).
There's not much difference for an average Daz user who's not modifying assets much (if at all) and sparsely moves outside of placing premade models within the scene often rendered with premade render settings.
You want original, you build or paint everything yourself.
How does this contribute to the AI argument?
Every country has its own laws and I don't think all would treat AI the same way.
So be strict about it and half of the world will walk over your corpse and you will be running to catch up years later.
Even TikTok banned AI Deepfakes so that should make it clear that just AI art as being technically good is not going to be enough to win the general public's trust or interest. Beauty is everywhere and not a lot of effort or money need be spent to notice it and enjoy it. If the public feels you are playing them, then they will eventually reject being played and legislate the specifics of that rejection if need be.
Absolutely. I was ignoring the focus a little bit here. Though i don't only mean legalities, i also mean structure and consequences, but that was off focuse already. Maybe using DAZ assets makes you like a (master) craftsman/artist with people working for you, while the generators are more like a wizard apprentice, where you never know ahead of time, if it will go haywire somehow. It seems to be different type of skills used, even for simple composition. E.g. in the prompt, you try to not let it render like Mickey Mouse or whatever, or you'll have to come from another angle language-wise. However that's a totally different approach, where you're sent on a journey rather. "Somehow AI" may actually bring us closer to the master craftsman eventually, but i don't see how current generators are there. A combined approach like FireFly may be closer than the magic general thing.
Freedom also doesn't mean that you always make use of it, though. So with DAZ you have different (more?) freedom and more control over the result, while in the simpliistic asset use, you'd be confined to what you've bought (and what you make of it).The simplistic DAZ user will "always" be able to reuse the assets they've bought, while the cloud version of a generator may change the price, remove possibilities, change style, change language models, etc. So the DAZ user may have a choice of what to buy and construct stuff from, even the simplistic one, for building their library. The generator person could "learn prompt", but that may all change drastically the next day, so you end up with learning "useless programming languages". That's not like jumping onto a train...
The big players all come up with something now+-then, maybe Mozilla too, and you know who next.
These models will need elaboration on what they've been built on, how they turn out to perform (usability, control, magic, results, something) and so on. So i would give it some time to get tested, actually. Maybe one will be better than the others, again and again, taking turns, or maybe each will be better at something, with tradeoffs and so on. User hardware also means different business decisions, or maybe means extra control software for licensing. Then the competitors will reconsider and probably adapt. Who knows. But for all those "centralized" models, i.e. cloud-only, there may be random changes coming.
We're damned if we do, and damned if we don't.
I'm a pretty private person (thanks to how I was raised, and due to life experiences), therefore (most of) my art is private. So even if I'm using AI for private work, I'm still an awful person for using it anyway.
Although I should start selling generations. After all, what's that saying... "better to be hung for a sheep as a lamb". Because all AI users are as bad as each other (although I don't believe in putting my work/generations behind a paywall, or selling off what could be another person's work).
Anyway, that's how I've started to feel.
The ethics case against Adobe is overstated, IMHO. See Wag the Dog. The Dustin Hoffman character is contributing. The 'pageant' does not get produced without him. He is not stealing from the Kirsten Dunst character, who is the fake peasant girl carrying the chips bag they use CGI to replace with a cat.
So Adobe is being sued despite its claims about having the rights to the data used for its Firefly artificial intelligence. Being sued and being unethical are two different things. Lawsuits are a given, and are just as often a sign of success (dividing up the goodies) as a sign of deep problems. Most of the great Motown performers sued Motown founder Berry Gordy at one point or another, and he countersued some of the writers. And there were lawsuits between Gordy and newspapers. And there were lawsuits between Gordy and.... In my opinion, Berry Gordy was neither all good nor all bad, but the world would have had a lot less great music without him, not a lot more. I'm sure plenty of people disagree, including some of the Motown performers and writers.
In my opinion, Adobe at least is making an effort to train its art AI with images and data it believes it has the legal rights to. I put the lawsuit of the stock contributers against Adobe in the same category as the lawsuits between the Motown performers and Berry Gordy. As it stands, I see no ethical problem with Firefly; stock contributors had a contract and their lawsuit against Adobe is about divvying up the spoils. The courts will settle who gets what according to how similar clauses have been settled many times before. Yes, the Adobe stock contributors consented to something in their agreement - and it sure looks generic - but what exactly did they agree to? The courts will decide. And the artists suing Adobe have professional lawyers with a specialty in this kind of law, just like the lawyers who represented Motown artists were professional when they sued Berry Gordy of Motown.
I provide a link to one article on one set of lawsuits involving Berry Gordy and Motown. The interested can easily find many more.
https://variety.com/2001/biz/news/gordy-stop-with-lawsuits-1117797798/
I've had no dealings with Gordy; therefore, I cannot speak to his business ethics (although the bar is exceedingly low for record executives), but I have had dealings with Adobe and consequently would believe any charges that anyone leveled.
Is a photographer not an artist? They don't create any of their own assets...but most do heavy amounts of postwork. The same as a lot of Daz artists. They create their own lighting and direct the posing of their subjects to convey whatever imagery they are trying to convey. The same as a lot of Daz artists. The only thing they don't create are the subjects themselves as well as any clothing their subjects might be wearing. The same as a lot of Daz artists. They need to understand the concepts of light and shadow, composition, and story-telling. The same as a lot of Daz artists. The same as any artist anywhere.
Yes, there are those who use Daz to render images of models and sets right out of the box with pre-made poses and pre-made lighting. Render-in-a-box is the term I think. But then there's a whole gamut of customization that can be done to make gorgeous pieces of art. Customized posing and lighting and tweaking everything to convey whatever story and/or imagery the artist is trying to convey. Plus a whole other gamut of postwork. Which is a completely different kettle of fish.
Good artistic photographers don't do huge amounts of postwork. Most of it comes from the lenses and their knowledge of light. They also work in physical world.
Color and contrast correction is also postwork...and most photographers do that at a minimum...even if it's with chemicals in a darkroom. Most digital photographers also sharpen their images and sometimes correct for things like chromatic aberration. And whether they work in the real world or not...my point is that they don't create anything they photograph. The artistry comes from the type of image they take and the story they create with said image. Their technique.
Two people stand side-by-side with the same exact camera and tools. They're taking a picture of a rock. One stands there and just snaps the picture with the camera set on 'Auto'. The other actually thinks about the rock and frames it artistically...laying down on the ground to get a perfect angle, capture the perfect lighting with the way the sun reflects off the surface of the rock. The camera is set on 'Manual' so they have to think about what shutter speed and f-stop they want to use...how shallow they want their depth of field to be. They want to draw focus to the bits of sand stuck in a specific place, so they decide to go wide open. Nice and shallow. Neither created the rock. But one is an artist and the other is not. The same concept can be applied to creating artwork with 3D assets.
I feel like lots of posts start going into double standard territory.
A person modifying renders is creative, so is a person modifying photos... But not a person modifying AI output?
I don't care about AI generated images, but it does look like some kind of bias is present. Legalities aside, because the AI might not be using "legal" dataset, but... What if a render artist or a photographer retouches their output in, let's say... 20%... But an AI user does it for 30% of the image?
you pretty much have to post work AI images
unless you like distorted pupils extra fingers, hands, and other horrors
in fact they are great for practicing your photo editing skills on, I am learning from that alone
Creativity isn't bound by much. Often being constrained to a set of assets, or tools, thinking of some crayons or maybe dirt and water, or wood and stone (...), leads to people becoming creative. The text-to-image suggests "everything" which seems the opposite of "confined", though of course it is confined in several ways. But look at it this way:
- Children paint something. For many children many parts of or whole images look alike. But when it comes down to using different means of creating an image, different tools, or the absence of standard tools, we'll likely see a relative abundance of variation. (In practice, if all the places where children are kept in the dozens, use the same tools, who knows, maybe it'll look similar-ish everywhere, more of the time.)
- Using simple words to describe what is to be drawn, with the same cloud service, ... "Imagine" how much variation there will be for the runaway creative person at the very young age. Of course, with humans, breaking out of whatever is a thing too, so maybe we get more diligent with using different language and try stuff early on, and maybe we learn our language from those generators :). Some people have learned some of their first bits of english from computer games, though arguably, language models are better at language than the "Space Quest I" interpreter, but the odd ways to talk to current image generators are of course leading nowhere, language-wise. If it became a thing, i would enforce legislation to make those things work on an english version of universal language, so the children "learning to paint" might also be able to talk to the aliens that have to rescue us later on.
I think i got off tracks a little bit here...
but i also think that this excursion also hints at more aspects of creativity. The photography discussion is as old as photography and it's already been answered, though i would always reconsider if it's too strict on "what's art". Even science or say: theory isn't stepping on the spot, and views change with artists doing new things. In essence there isn't even a question about if you can do art with ai generators, or with a wooden stick. In most discussions, i'd assume, people are talking about different things, pre- and side-conditions, most of the time. It's easy on the internet.
You could ask, if you're an artist now, because the lush triplets you throw at the prompt output something, that looks like art, or even look great. That of course... depends.
Just recently i was "recording" a blackbird frantically picking the eyes of a roadkill bird. I thought that was a blackbird too.
Now i thought i was recording video, but i wasn't, i was just switching between high res and normal aperture mode, at least that's what i concluded when looking at the results. I only have a few not so sharp photos of the incident. Had it been a UFO...
Meaning it's probably like "AI art" - you have to do the art, the ai doesn't. If you think it's "making the art", you're like me, who thought the camera was recording.
That perhaps?