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
I still have 44 years before hitting three figures, so... Maybe I will see the culmination of the personal quest that started 23 years ago with trying to get sound in and out of the computer
Yes the HAL9000 AI vs the Google Car AI. Or "Strong AI" vs "Weak AI" or "Generalist AI" vs "Specialist AI".
The HAL9000 AI is (at least for now) science fiction and really doesn't have much use. AI as we see it now is specialist AI. I don't care if my self driving car can compose a Shakespeare sonnet. I just need it to be the best at driving. Creating a HAL9000 style AI just to drive a car would be overkill and if AI ever gets to the point that it is considered a sapient being then shackling it to the task of driving my car would be tantamount to slavery. But a specialist AI dedicated to being the best driver possible is not that hard to achieve and improve upon and over the next decade we'll be seeing specialist AIs taking over many industries.
I love modeling. It's a fun mental exercise. We're already seeing AI take over some tedious aspects of modeling. Retopology for example. It's very easy to sculpt a neat looking model but the tedious part is converting that high polygon sculpture into a lower polygon mesh that is easier to work with. Auto-Retopo is a great tool. A lot of times when I sculpt something in 3DCoat I'll let the AI retopo it and then go in and clean up the areas where it doesn't quite get it right. This is a lot faster than manual retopo. Each new version of 3DCoat the retopo AI is improved upon. In a few more generations I probably won't even have to think about retopo, it'll just be something the computer does on its own.
Animation will probably also be automated more. Basically you get a learning AI to watch a whole lot of animation and learn about dramatic movement. We'll be able to say things like, character A walks from here to here, turns and says "you cad, you scoundrel!" Then it animates it and we refine the animation the same way a director refines the actions of an actor. We'll be able to say things like "make the motions angry, but still a little bit cartoonish" and the computer will understand angry and cartoonish and deliver us an animation which we, as the director will refine more "no no, Bill Murray cartoonish... not Dan Aykroyd cartoonish" and the AI will know what that means because it's been trained by watching every Bill Murray film and every Dan Aykroyd film and it knows how to mimic their acting choices.
I'd estimate that traditional key-frame animation has maybe a decade and a half left in it at tops and we'll probably have simple AI animation apps available on the consumer market in the next 5 years. Indeed I've been watching YouTube experiments in AI animation and they're already looking pretty promising if all you're using animation for is Vtuber stuff. Lighting will be another task given to AI probably combined with physically modeled lighting simulations rather than temp/hue/lumin lighting.
There will still be human involvement but the human will be the boss directing specialist AI underlings.
I think if an artificial lifeform style AI comes into creation it'll probably be by accident. One of the things we've learned from self learning machines is once they learn one task, they learn other tasks quicker, easier, and better. People will probably network a bunch of specialist AIs together to assist in machine learning and before we know it an artificial lifeform is born and then probably a singularity.
All I know is I tried Artbreeder which uses AI to radically enhance images... I haven't rendered a single portrait since or used artbreeder again.
For me? I just couldn't figure out what the point was anymore. Just throw any seriously crappy picture/render into an AI editor and let it turn it into something amazing and 10 times better than I could do on my own with all my efforts and skills.
:/
I'm moving on to basket weaving.
Oh, I'm sure AI will muck that up sooner or later. But I'm banking on later.
:)
I never heard of Artbreeder before. Looks like it would be an awesome inspiration generator though.
Haha
AI is not revealing anything new, most of the time it's just revealing what we didn't realize so I wouldn't be too dismissive of it too quickly already I hear the same tune every time computer automation making inroads into some niche's territory: "Oh computers can't do as good as people. There is something missing." But computers can do as good as people and that something missing is the jobs formerly held by people instead of computers. It can and is being abused posing questions and making purported moral judgements where there are no such judgements to be made. That's the fault of the designer, not the mathematics.
The good thing is I've yet to see someone cry because a computer fried on the job and won't booot anymore or a computer need to file a OSHA claim because of a work injury.
If AI can write stories etc. as good as Shakespeare or even not as good as Shakespeare and can explain the internal meaning, then it means it can think! Until it can think, it will never become the best at driving.
Current AI is just set of rules.
This is the reason I gave up playing chess decades ago.
Well the thing with AI is it doesn't have to be the best. It just has to be better than meat. Self driving cars have a much better driving record than meat driven cars. AI surgeons have a much better record than meat surgeons do. We might some day have AIs that can generate creative content better than a human can but I think that's going to be more of an academic exercise than something that will be generating all our entertainment. Humans enjoy being creative and always have and there's no way that'll change. Most humans don't enjoy drudgery however and AI doesn't care about how tedious and soul numbing a task is because AI doesn't get bored. We'll probably see AI excelling at being a creative assistant but not being the creator.
Again, can we please dial back the dicussion of AI in general and focus on the specific aspect here.
@nonesuch00
I really want to say that;s hyperbole, but I'm afraid we can't...
Neural Nets are not actually new, they've been around since the '60s and mimmick the actual functioning of a biological brain. A "neurode" is just a mathematical model of an actual neuron, and so "AI" is only "artificial" in the sense that engineers new it would be better to base it on silicon. What made "Deep Learning" possible were three things: Exponentially more powerful processors, exponentially larger data storage, and exponentially larger training sets.
But to me the truly scary thing is that unlike, say, an Expert System, a Neural Net could give you an answer, but it did not offer a way to request the Neural Net's reasoning. That is, it could tell you the What, but not the Why. Given that, we really have no way to know what a Nerual Net is really thinking, we just know its response to input and nothing about what happens in between. There's no way to know if the Neural Net isn't a complete sociopath that smiles at you while contemplating murdering your family at a more opportune moment. Nerual Nets fell out of favor in a period call the "AI Winter", where there was no funding, because they lacked this kind of traceability.
And think... the larget Neural Nets we can build are on the order of the complexity of a fly's brain. And we already can't understand what is really going on, presumably you would need something even more complex to understand it. What hope will we have when we've got a NN comparable to a human brain? How far beyond us will be the ability to understand the reasoning of something like that?
And what happens when (and I'm absolutely sure this is the very first thing engineers are going to do) we have such a NN and we tell it to design a NN whose only purpose is to design a Neural Net whose only purpose is to design a Neural Net, ... that is smarter than itself? We will soon have machines that far, far outstrip our cognitive ability to understand anything that they do.
That, to me, is even scarier than Heat Death of the Universe because it's going to happen a heck of a lot sooner.
@PerttiA Elon Musk thought the same thing and made a billion dollar bet that you are right.
I read an article recently that said the recent fatality (R.I.P.) involved a sort of perfect storm — I’m paraphrasing — of unexpected factors like the cyclist being (at least briefly) a pedestrian at a crosswalk.
I went to one of the big stores one time, in a big mall in my major center and I tried the glasses etc... but whatever was supposed to capture my attention did not seem to do it for me. There was also a time when all I had was a little black and white TV set and the local public/educational station ran a spot where somebody had a broadcast-side scheme to send out a stereo signal, I think involving rapid left-right switching like some people like to do with animated GIF's. I thought the brief demo had promise, particularly since it didn't require me to do anything really special.
Remember carrying cassette tapes to use outdoors with your portable tape player? I'm also reminded of a recent documentary "The First Film" where technically speaking the guy (Louis Le Prince) not only came up with a camera and a player, but also the idea of a commercial theater; IMO about the only thing missing was the buttered popcorn!
Just because something doesn't make it in the marketplace in its initial manifestation doesn't mean it wasn't cool/enjoyable/useful. My "View-Master" 35mm camera may be be gathering dust but I still want a Kula Bebe! (Made by a woman in Iceland - look it up.)
I liked Frank Herbert and "Dune"... I often find myself reflecting on Cogitor Kwinna, Omnius and the independent robot Eramus.
Oh wow. This could be a big time saver for rendering in the future. Absolutely amazing some of the tools that will soon be available.
what a time to be alive
Ai would suck the fun of the reason why you make art, so myself as a self describes artist point of view who likes to create things on my own, whether succeed or fail.
I ask y'all why bother making 3d art or investing in 3d content at all. seriously if Ai is the next thing in the revolution of things . then just buy a real nice Ai system and tell it to make art you want & your done. no thought required. nor is there any reason for forums or a rt stores. well maybe a few forums on how to get control of your ai system once its deemed you to stupid to live.( ya I watched all the terminator movies.)
So if Ai is the next big thing, where is the fun in creating art if you have a computer do it for you,? Personally I see no use for Ai in Daz studio or for any other 3d software where the artist who use those app would like to think and create for themselves.. Ai IMHO would just suck the fun out of everything. it be like turning on your computer and telling your computer to play forth-nite for you & you just sit and watch.. Boring!
Cool, I think similar is integrated into UE5 when it gets released.
Well, what passes for AI has its uses with the kind of frame interpolation we've seen in examples. You have to be careful with it, but it's conceptually the same as the producers of series like The Simpsons hiring inbetweeners in low wage countries to handle the boring part of doing cell animation.
As far as the creative part, it's like how some people get used to drinking instant coffee and iced tea from a mix and watching porn and eventually come to prefer them over the real thing. Whatcha gonna do?
Ai maybe useful for mass production usages saving time and labor, like they use robots for assembly line production . But as far as a hobby person goes I think Ai would kind of defeat the reason for having the 3d hobby in the first if you going to just sit back an watch the computer make a art creation for you. where would the fun in that be? It will only give more credit to the people that already say using daz is not making art because of premade content. Wait until Ai drops onto the scene. Artist? We don't need no stinking artist. we have robo-brush 3000..HaHa
There are already some pretty amazing experimental apps:
https://www.youtube.com/watch?v=jMZqxfTls-0
https://www.youtube.com/watch?v=JlTEA0VevSY&t=44s
https://www.youtube.com/watch?v=4MfG9CDufPA
The ENHANCE one that Ghastlycomics posted is impressive!
the portrait painters probably said that about photography
Ai is going to need a robot to paint a portrait . and i would properly pay to watch a robot to do that.
The people who made shadow puppets on cave walls probably said that about painting.
Hey I am pretty good at shadow puppets. I can make like 3 of them now .
I don't like the idea of this at all. I want to be in control, not just giving instructions and hoping they get carried out like I want. But perhaps I'm not typical. When I was a programmer I liked assembly language, I was annoyed when they said I had to use "high level languages" instead. And I never followed the typical path of IT people becoming a manager. I have no skill or interest at all in management.
I expect there will be a range of opinions, from the craftsman style of person who wants to make things themsleves to the mass production approach. Maybe professionals who need to do things efficiently will prefer automated AI systems.
I started off in 3D playing with some early modelling and rendering programs. In a way the whole Poser/Daz Studio route has been a bit of a diversion for me, a way of getting quick results. I keep intending to start modelling stuff again, maybe I'll make time for that soon.
https://arxiv.org/pdf/2010.05562.pdf
A couple of other examples of some awesome AI imaging research that will probably apply to our area:
https://www.youtube.com/watch?v=gg0F5JjKmhA
https://www.youtube.com/watch?v=P0fMwA3X5KI
And yet another cool example:
https://www.youtube.com/watch?v=u7kQ5lNfUfg