HAS THE CLOCK STARTED TICKING FOR DAZ?
nelsonsmith
Posts: 1,334
I'm pretty sure all of you have been if not intently, occassionally been made aware of the rapid advancement of AI when it comes to art. For me I like 3D modeling and DAZ3D because at least when it comes to 3D artwork I do happen to like the process, but I can completely understand how hobbyists and others that are more interested in the end result would easily and quickly gravitate to AI, since they simply want to create the images they see in their head. Same for those who were chasing photorealism; now it's completely achievable.
So the question is what percentage of DAZ users that were dabbling around with the program can achieve what they actually wanted with AI, and perhaps with considerably less cost? A lot of people might think it doesn't matter, but I used to be really into the Super 8 film scene, which the advent of video killed almost overnight eventhough video at the time wasn't even comprable to the quality one could achieve with super 8 or 16 mm; the general public could pick up a video camera and shoot without the need for learning F-stops and exposures. And DAZ is really kind of in the same position. The learning curve to actually use DAZ well, or at a professional level is pretty steep and the info to do so is so haphazard it's a chore to map out any cirriculum to master it. I've always said that that was a major turn off to potential users of the program, and now it's probably the reason a lot of people are also going to move on to something else.
Anyway, it's going to be interesting to see what DAZ is going to look like, and where it's going to be in 5 years; especially since there was no glue on the seat of their pants when it came to incorporating their own AI version.
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
I find AI art is like the advent of TV dinners. They both might have their place in society, but no one is going to accuse the one of being really good food, or the other being real art...
I think Daz needs to go there. Not necessarily point and click - there are many AI-assisted tools they could integrate that would be more analogous to the difference between mixing your own paints and buying some at the art store.
This has been asked before and just tends to get contentious. In any event, there is a forum for discussing Daz AI Studio here https://www.daz3d.com/forums/categories/daz-ai-studio
It is true that the Night Cafe I have the entry level sub to is a beehive of activity compared to the DAZ Gallery but the fine tuned control an artist wants to have for messaging in that AI art is just not even close to happening. Not even close.
AI ... it really isn't. Not the way we used to use that term in my day (back in the late Cretaceous). These engines associate impressively, but they understand not at all. They include only part of what we used to include in the AI concept. Which means --
I quite agree that AI as we used to mean it would blow Daz (and a great many other enterprises) clear away. But that's not what we have.
"AI" as we currently have it will no doubt form a) a hobby and b) a source of fast and easy illustration, for applications where you only need a single piece of conceptual art. Say, blog post illos -- I can readily imagine that some people will want to prompt pictorial engines instead of browsing pic sites.
I can also readily imagine that, after I have a piece of art about 93% done, I might want to see what special effects I could get an engine to add for me. An extra versatile tool in the box. Sometimes it will produce effects I didn't imagine, with my existing graphics tools, that I like. Great!
I already play around with a pictorial application that has somewhat similar features, except in a more limited domain. It's J-Wildfire, and it makes flame fractals. Chance-formed fractals can be as ugly or even downright evil-looking as a welter of diseased space fungus spores. Or they can be sweeping, beautiful, intricate. There's a lot of randomization to it. I can control some aspects of the pictures produced, and others keep twisting away in ways I didn't expect ... which might suit, or not. I've been known to engage in the time-consuming and impressively silly endeavor of coercing J-Wildfire's mathematical formulas, over multiple layers, into producing images of obviously physical science fictional objects instead of strange patterns of geometry and plasma. (Silly, because I could mostly draw the things faster, if I started with those particular objects in mind.)
My experiments with 'AI'-prompt drawings produced much the same sensation as playing with J-Wildfire, except that the failures can be not merely unlovely but hideously deformed. Five-legged horses are the pretty failures. The bad ones ... eeeyeccch. I'm not sure I'll do much of that.
My judgement.is that I will want to use these engines as an adjunct. Some people will probably use them to produce basic illustrations where they don't have any strong desire that a particular thing should be shown in a particular way, so long as they get something to illustrate the concept. Some people will put effort into getting some very good-looking works out of them (I've seen some).
This kind of 'persuade the engine' work feels different to me than direct artwork, and yields a different sort of result. I can imagine adding it to the mix, but it will never replace more direct controllable art for me. The fundamental thing that drives me to create pictures in any medium is to fix the partly sharp, partly vague, half-imagined shapes and colors in my mind's eye into definite, direct, and striking form: to bring the whole vision into lasting form. The algorithms will produce some vision: but not the one that teases me.
Also:
https://www.youtube.com/watch?v=NcH7fHtqGYM
"Scientists warn of AI Collapse" -- A very brief presentation by Dr. Sabine Hossenfelder (a physicist).
A ~6 minute discussion of the fact that we are now feeding AI-generated data to AI generators. What will come of it? And what are we going to do about it?
I am reminded of the Larry Niven short story "Convergent Series."
The clock hasn't been wound up yet.
I will be concerned when the various AI engines can take a two-page detailed narrative description and faithfully render the result down to the last detail.
My work is my work and I have no interest in ai enhancement. So someone's gonna say, well you used some stock assets and I'm gonna say, I put enough time and effort in, make enough modifications by the time I'm done with it, it's mine anyway - it's not really the stock items any more. Perhaps it's a collaboration of sorts, and that's pretty cool too.
I would also add, ai is just as much a source of electricity use and thermal pollution as Cryptomining, NFTs, and other such technologies. Your ai leaders are busy building huge new computing hubs to feed the beast. That's not being contentious. That's a simple fact of life that doesn't get near enough mention.
There are alternative, more power efficient architectures to x86. Power consumption is probably the #1 bottleneck we're going to hit with AI. I wouldn't worry too much about AI indirectly cooking us, we're going to switch to RISC sooner or later.
I was explaining my new hobby here to my mother the other day. She was curious, so I'd sent her the render I did for the April Fool's contest as an example, and then explained where I got everythng, and what I had to change about it to make it work. (And how many things I discovered DIDN'T work in the process.)
"Oh," she said. "So you're a 3D art chef."
And I just sat there blinking for a moment because yeah, that's the perfect analogy. Maybe I didn't grow any of the ingredients myself. Maybe I even paid for some really good sauce out of the jar, or whatever. But I still had the vision to put everything together, and to find substitutions from what I had on hand when what I wanted wasn't available.
So add me on the train of those not interested in AI. Yeah, the learning curve here is steep, but the people on the forums are kind and generous with their hard-won knowledge, and the other day some tutorials a kind soul linked me gave me the ability to modify some stuff into what I wanted it to be, and I strutted around like a conquoring army for the rest of the day.
AI will never, ever give me that rush or feeling of accomplishment. And I am very much not willing to give it up.
I get the best of both worlds. I still set up a scene in DS and render a quality image, then use AI to enhance it. Nine times out of ten I get a more realistic image. For me, AI is more for postwork than anything else at this point.
Another neat use for AI is to use a DAZ render as an image prompt, which means the AI creates a whole new image based off the original render, the look, the pose, etc.
I like to use AI to generate images, it"s fun, but I am very aware that all I am doing is just typing text or using an image prompt and hitting the generate button a thousand times, very little creative input from a user other than deciding what to type and what checkpoints or LoRAs to use depending on the UI
The Vorlon ambassador says, "The avalanche has already started. It is too late for the pebbles to vote."
https://time.com/6987773/ai-data-centers-energy-usage-climate-change/
https://www.datacenterknowledge.com/ai-data-centers/bitcoin-miners-pivot-to-data-center-operations-amid-ai-boom
https://www.forbes.com/sites/cindygordon/2024/03/31/microsoft-and-openai-partnering-on-stargate-a-100b-us-data-center/
Need I go on? That's all I will say about it.
Clock's been ticking for awhile. Once the codgers pack up, the graffitti will become legible. As people mentally outsource more completely, individual creativity will become a derided hypothesis.
@SilverGirl "Oh," she said. "So you're a 3D art chef." Brilliant! And thanks. I enjoyed your reply.
Glad it could bring a smile. :)
What's your point? Nowhere did I claim that data centers aren't power-hungry molochs. I'm saying that they soon will be outdated and replaced by more power-efficient operations, so there's no need to be pessimistic.
The change from Super 8 to video cannot be compared to moving from 3D to AI assisted 2D.
Moving from using DS to create the scenes to using AI to create 2D scenes, is like playing Skyrim or just looking at snapshots from the game.
AI would be great to reduce render times in DAZ, or to melt a seperately rendered prop into an existing image.
There have been so many far superior artists creating art out there since the beginning of DAZ compared to the average daz user here, and still we all have continued to buy and create for ourselves, for our friends and family, or to show here to other fellow daz users - so even if ai should develop into a make art button someday, there should always be people who wish to be creative themselves to a further degree, even if it can't hold up compared to the art of other talented people or machines. Many people want to create for themselves. That will always be supporting the classic daz workflow. Life is not always about being the best or saving time and reaching a higher level of productivity.
I work in an engineering environment, and the owner of the company is trying to persuade me to use all sorts of AI.
My concern, illustrated by the AI "Well that didn't go as planned" bloopers thread, is that the AI has an understanding of a sort what's being asked and reproduces that understanding. Unfortunately it has no conceptual understanding of the objects it seems to be reproducing. By this I mean 'What is the skeleton inside the human and why does it limit the limb positions and limb numbers?'. It's that incomplete conceptual understanding (if there is any understanding at all) that is dangerous to start using in real life. Things work (or not) for a reason, and AI at the moment seems to have no understanding of that. What's the saying? "A little knowledge is a dangerous thing". Then combine that with a very little level of understanding, give it to a bunch of people with no understanding themselves, and you can end up with an entirely foreseeable lethal event.
I do think it'll get better, but I'm not convinced that the real understanding will come in the Large Language Models currently in vogue. There is a depth of conceptual understanding needed that doesn't seem to be currently developed.
At the moment, with my work, I don't think AI is applicable, but I'm watching it. Same as I'm watching 3D printing. The3d printing materials are getting closer to being long term usable for things other than toys. The same will happen with AI.
Regards,
Richard
In many fields AI can be useful as an additional tool if actual professionals can give it the right inputs / orientations and challenge the outputs, but it couldn't do the work alone.
Agree 100%.
I am going to use photography as an example of what I think will happen with AI.
Photography was both hobby and semi-profession for me ... and mostly an art form. Do not think point and shoot camera, or even an SLR with interchangeable lenses... think a large box camera with bellows (like Ansel Adams used) ... and I did my own darkroom work for black and white.
Still to this day, there are things you could / can do with chemical based photography you cannot do with digital ... but those things are really pushing the art medium to its limits.
For the average person taking snapshots and the the average professional trying to be quick, efficient, and profitable ... there is no doubt that digital photography is the way to go. And, yes, with tools like Photoshop it can do things that chemical based photography could never do.
So, digital photography took over (very very quickly) (and yes I too use it) ... and chemical based photography has been relegated to very specific tasks or artwork. Probably much the same will happen with AI.
EDIT: I have been playing with the Daz Studio AI ... even with it's limited abilities compared to other prgorams available, I can come up with quite usable images for my powerpoint programs very quickly and easily ... but if I am to go to print in a book, I would want to revert back to Daz, using the AI image as a base idea, and make a much better and more detailed scene. [ I cannot say what other programs would do, as I have not used them ].
And I guess I'd say this seems to be working out to -- if your reason for using something is fundamentally utilitarian, and an easier, faster, cheaper method of mostly doing the same thing presents itself, you'll probably switch. If an easier, faster, cheaper method of satisfying the same kind of aesthetic impulse, or that produces much the same sort of satisfaction, you'll probably switch.
For me, typing at Midjourney produced some images I might use as web furniture. But it didn't satisfy the same aesthetic and creative impulse as 3D. It gives less control, not more. I want more control, not less. So if I drift away from this section of the 3D market, it'll likely be because I'm making more of my own 3D items.
I have the impression that there's some professional use of Daz models and renderings, but most of the customers are hobbyists. And to me, typing at AI generators feels like a different hobby.
So this is isn't the market I expect to suffer greatly from AI image generators, unless I'm wrong about why most Daz and Poser customers buy. But some of the stock 2D graphics vendors are likely to take a hit.
I enjoy the challenges of creating 3D clothing and textures much more than I do the challenges of finding the right combination of keyword to get the results I want. Some people have mastered the AI universe very well, others not so much. I'm one of the "not so much" crowd. LOL
The way AI is trained - The way training data is harvested and processed and regurgitated - also remains a problem. It must be said as often as humanly possible. Now I've said it and I can leave this discussion to its fate. Do have fun now.
Except this question, isn't really about Daz AI Studio. All you have to do is look at any of the work being done with Midjourney and whether Daz incorporates and embraces AI or not doesn't matter, the fact is the business model Daz is using now, is nearing an end. Why wait for a PA to create something, let alone pay a PA for anything when you can get what you want with a prompt? The question isn't if DAZ is going to become a niche tool, it's going to be how small is that niche going to become. Or at least that's the way I see it. It's rather sad, because DAZ 'was' heading in the direction it needed to be going, it was just moving too slowly, and now it's going to be extremely difficult to gain a larger customer base when you've got the major competition that is now out there.
And that could have been DAZ's save card. AI as simple as people seem to think it is, is not that easy to get the professional results that I've seen. Also since no one really owns the software, but rather I guess they share it, it's not that fast. The problem is I can find dozens of online courses on how to use the major AI programs, Udemy ahas several "Masterclass" on AI art, as does Wingfox ( which is primarily a site for traditional artists ). I can't find one similarly structured class for DAZ on any of these educational sites.
>> I can't find one similarly structured class for DAZ on any of these educational sites.
Being that it's still in beta you probably won't find any comprehensive tutorials until it's done
EDIT: I've recently retired from documentation work because at my age (septuagenerian), the old brain ain't as sharp as it used to be. After a two year hiatus from content creation, it's taken me nearly a year to get up to speed on the stuff that used to come like second nature. I gave AI a try out of frustration, but my brain just doesn't want to go there. So I'll stick with what I love to do and it will come eventually. 8-)
The problem with the kind of ai that (has to be?) used for image generation, you are bound to change mindset and perception, in order for it to really replace stuff (i.e. you change, your work changes). The other point is, what the customers of the image-creators prefer to use (revenue, jobs). Since Prompts can yield nice illustrations, with inpainting and modification thereof, plus cheating, there certainly will be powerful tools. It does consume a lot of power, and the large models are very expensive to train. With cloud-based services you will never know, when they will shut down, or change significantly. We've had those "this is better for you now" moments with various kinds of software, not just cloud applications, but also browsers and operating systems. So far, technology somehow advances, but benefits are clearly stalling, in comparison to (theoretical) capabilities. So maybe you have a tool then, maybe it works for a year, maybe for ten. Maybe you need to re-learn prompt-engineering tomorrow, and maybe your images get blocked for hate-speech or whatever. Maybe it'll be so much better the next day, but you can redo all previous projects due to inaccessability of what you did before (there are tools for adjusting to style, though). Nobody knows.
That's one side of the coin. There are already promising approaches of making networks smaller, for more efficient training, and other ways to train, like "dreaming". There is near-endless amounts of topologies to try out and there will be further advances. Just like with the space shuttle. No idea what this means for DAZ, but i think i got some part of my point through.
Or did you mean this ticking clock: https://thebulletin.org/doomsday-clock/
To be fair, alot of the courses on AI are targeted at the lay people who get into Generative AI due to the low cost of entry.
So I would estimate that the number of people ,from the general public, using generative already far exceeds the number of people who have invested in a modern PC with a NVIDIA GPU to run Daz studio.