Future of 3d modeling and texturing
polycounter lvl 5
I have been learning 3d since last few years and i am very passionate about that. Recently i saw a video on Youtube from BlenderGuru, he had an interview with the Blizzard's Environment Artist Michael Vicente. He talked about the AI replacing most of the 3d jobs in the industry. I am pretty worried that if it is going to replace most of the 3d artists then whats the point of having an artist? and no artist would like to work as a technical person cleaning up stuff for the AI and photogrammetry . Please can someone clear this to me.
Replies
Into your diffuse textures?
With your hands?
Shadows!?
Now shadows are done by the software!!! What should we artist do now?? Just painting pure colors??? :´(
Things will always change but there will always be room for artist's decisions.
However: be nice to your computer. One day they will be smarter then us and they will kill us all if we aren't nice to them now
One good metaphor I've heard is AI is like having a massive army of toddlers at your command. They can do some things very fast, but they're still completely incapable of doing other things, and controlling exactly what they do is very, very hard.
AI will probably decrease the number of artists required on a project. But that's not necessarily a bad thing for artists; if you can bring down the cost of graphically rich games, more graphically rich games are going to get made. They'll still be need for artists, and we'll get to work on more interesting projects, and spend more of our time on real creative work and less on grunt work.
Check out this video, from the founder of one of the most promising startups using AI for game art production:
https://www.youtube.com/watch?v=9FAXAgRrOSE
The problem with AI taking over an artists job, is our job is continually evolving, and requiring more and more and more work.
As long as computers get more and more powerful, there is no limit to how far we could push our visuals.
If you told me 15 years ago, I'd be counting eyelashes, and doing 150+ FACS blendshapes, and spending several months grooming hair, and actually appyling the "proper types of pores" to each area of the face, id tell you that would be improbable.. yet here we are.
If you told me we would be simulating cloth in a fashion design program, and it still wouldn't speed us up very much, I'd say you were out to lunch.
I agree with most of whats been said, we aren't all of a sudden loosing our jobs over night.
but, let me assure you, art is by no means some magical field that AI will never be able to replicate because of the human soul or something.
we are all getting replaced sooner or later.
That's all
I would argue the first one is what makes one think "wow this is an crazy amount of work and craftsmanship" but on the other hand I hardly see any creativity in it. Just let an simulation (not even AI) do it way better and way faster.
But making a great costume might be way easier (if we image the perfect tools) but would require something I would describe as artistic creativity.
But everyone can come up with 'something'. It's highly subjective and a lot harder to define in 'good' or 'bad'.
I - personally - think that picking the tools already defines a bit how much _Art_ you can bring into a piece of work. If you run a cloth simulation you get physically correct results. If you do it by hand you can do stuff that goes beyond that.
So I would draw a chart that defines creativity on one axis and physicality on another axis
And yes, this post got bigger then intended...
this happened when normal maps and highpoly workflows became more common in the ps3 era, especially in vancouver, a lot of people refused to learn new tools and adapt and then ended up leaving the industry after a huge round of layoffs and not being able to get a new job, because they couldn't demonstrate an up to date skillset.
it happened again this gen with the advent of pbr and the substance tools suite, anyone who is unfamiliar with those tools will struggle to get a new job if they have been coasting for the last few years and get blindsided by layoffs. I know quite a few artists who would be in a horrible situation if that happened and they are constantly saying "yea, I really need to learn substance...."
The same thing will happen to other industries, like the trucking/shipping industry. Anyone looking to become a truck driver these days is probably in for a rude awakening in the next 5-10 years if they don't take automated fleets of trucks into account. Mcdonalds workers are already getting made irrelevant by automated ordering kiosks - my own behavior adapted about a month after those showed up to the point where I NEVER go to order at the counter any more.
creative decision making and having a more in depth impact on overall player experience in roles such as design, world building, etc are going to be even more important skills as all the menial work gets automated. It's pretty inefficient to have humans hand placing rust leak decals under things, or hand placing little tufts of grass in the cracks of sidewalks for 8 hours a day.
If you see the changes coming, it is 100% on you to adapt to it to remain relevant (good companies will). Don't expect a company to train you to keep up, a lot of that is going to have to be learned outside your 9-5 hours. But as new tools and systems come into play, there is a HELL of a lot of opportunity for those that embrace them first and master them. look at all the initial substance designer gurus, they built huge followings, sold a ton of tutorials and were in extremely high demand for the first few years substance was out and becoming the new standard for materials.
evolution in the tools and industry is just another chance to create opportunity for yourself.
We were wondering if a very sloppily and shoddily made game would be "acceptable".
Imagine we scanned a character, then just ran Zremesh on it for the lowpoly.
Auto unwrap the UVs.
Imagine all environments were photogrammetry, and all props are scanned.
Funny thing is, with the power of computers today, that crappy looking character we made in under 30mins, would probably be better looking than MOST games made 10 years ago.
Maybe if we were to reframe the question it might yield a different result:
How much more power would we need for there to be a negligible difference between the sloppiest cleaned up auto-retopo'd scan were to compare to properly constructed and optimized lowpoly one?
In a production environment unless you're a concept artist or are directly involved in art direction / production design etc. you're an artisan not an artist - you are there to realise other people's ideas. This is very much within the realm of possibility for AI and machine learning.
The main reason it hasn't happened yet are because there are better ways to make money from AI research.
So many art styles out there to choose from, that don't require a giganto art team or a gazillion years.
https://polycount.com/discussion/211344/what-low-polygon-art-styles-do-you-enjoy
Its just as possible that at some point, AI will not only be faster and cheaper, but also more intuitive and creative as well.
There have already been "documented cases" of AI being "creative". When AlphaGo defeated the best human player in the world, Ke, he said the computer was not not only better then him, but regularly exploited tactics and solutions he could only dream of, or had the clarity of mind to perform maybe once or twice in his life.
the same goes for the AI that defeated the Pro players in Dota 2.
obviously we are still very early in this, and a game/sport is not the same as art, but I would definitely say it is creative in its own right.
I wouldn't be surprised if AI can crack the algorithms of creativity at some point.
Maybe human players might learn to cheat in that way quicker. But if learning to cheat is just a matter of trial and error, eventually AI would most definitely be doing that much faster. In that case, wouldn't you say that is AI being more creative than a human? I guess real creativity would be realizing the whole thing is a game and somehow like, escaping it entirely. But whatever the point is... Skynet is coming.
This provoked a wry smile, AI algorithms are already having their - it? artworks auctioned by Christie’s:
https://www.americanscientist.org/article/ai-is-blurring-the-definition-of-artist
...gimmicky at best, to be creative a soul is required.
for me the discussion is bigger then that, Its not about my creativity vs yours, or an AI for that matter.
my main point is only that AI will get better then us at some point, denying that feels very naive to me, what we do about it afterwards I have no idea.
perhaps human made things will increase in value simply for the fact they are "hand-made".
While this kinda stuff has been around for decades, what Nvidia has done is quite a big leap in terms of usability/production. The more they refine the AI (or somehow make it able to learn on its own), the better the results/output will be. I'm sure at some point there will be public tools where we can just scan through google images and generate an 80% ready production mesh from a single image.
I personally think AI and autonomous art are great for the future. They will force artists to be more creative, rather than 80% of the production workflow being mundane and cleaning up topology or some other rubbish. There are already some nice tools that make production-ready meshes instantly from high poly sculpts/concepts. Mike Pavlovich I would say is a pioneer in pushing this forward.
The only concern I have is that the standard of art is increasing higher each day. Stuff that was in AAA games 5 years ago no longer is a wow factor to most people today. Seemingly beginner artists also seem to be producing some insanely good quality work these days too, but that's attributed to the amount of teaching content available out there.
I've also seen a lot of stylized art on the same bar as Overwatch, be cranked out completely from a concept sculpt + procedurally generated textures from substance do 95% of the work.
So a greater emphasis on learning the art fundamentals will differentiate an average artist from a professional in the future. Once you can teach AI art fundamentals, which WILL happen, I'm not sure where we will go from there. The AI will be able to produce better quality artwork than we can imagine, considering that 1. It will have an infinite visual library of material and references from online and 2. It can crank these out like clockwork.
dont missunderstand me, Im not really scared about AI taking my job or anything. There will always be something to do, even at the point when AI can do all the things we can, and do it better then us humans will always find a way.
For me its all about being prepared. knowing AI will do EVERYTHING we can, but better, at some point is a reality that the quicker you except it, the better you can prepare for it.
I feel like I'm linking this somewhere around here every moth or so now, but I think this video essay really paints a good picture of the reality to come.
If anyone is either worried, or interested in AI's role in the future This is a must watch!
I am certain that AI in ~200-400 years will take over most of the jobs we know today, because there is no reason to stop this technological advancement as it brings exponential growth of productivity.
In the near future though, AI will not take over the creative industry. It rather starts to take over different sectors like the engineering sectors, research sectors and generally data driven sectors.
The reason is that AI research is based on data. And the more complex data is, the more difficult it is to implement such algorithms. Now compare the dataset of a 3D character: Rigged, animated, textured with for example Customer Data for an online shop like Amazon.
Now compare the complexity of implementing an algorithm which automatically chooses the best placement for rigging bones joints with an algorithm that choose the right advertisement for your customer. There are worlds between these use cases.
I am not saying AI will not find its way into the gaming industry, but it will be alot later than it will be in other fields.
20 years ago, it was said 3d would replace 2D animation... and look, There's more 2D Anime than ever.
10 years ago, or more, it was said we would be using really great computers, with excellent multicore CPUs... and look, we still have the same as 10 years (Intel, 8 threads and 14nm). We won't have PCIE4 in the newest intel platforms, nor DDR5 memories. All Delayed, like happend with DDR4 and too many things. And it's ironic, a super mega duper AMD Ryzen 3900x is still not enough. Actual technology is stucked, and the improvement margin is minimal. And where are those Hard Drives of 100+TB announced back in 2010? Seagate claims they will have 50TBs in 2026... They bring out "Specter and Meltdown" just to force us to buy new cpus+mobos+ram, it's like a forced scheduled obsolescence.
But well, we always can dream about the future and its miracles, in 20 years we could find a way to inmortality. Days ago, i readed an article about all this, it was related to the Anime Film AKIRA, and how things were not as "they expected".
I also would like to see that "stylized art on the same bar as Overwatch"
Its nothing AI couldnt achieve. Take the process of finding ideas for example: Humans have a lot of emotional memories and general knowledge. When we find ideas, we basically connect random parts of our memory to create something new. We make a new connection between memories and form something new out of it, which most of the times means that idea is not actually new, but is combined of many components that already existed. AI or computer technology in general does that very, very well. Alot faster and more efficiently than the human brain can. Especially since most of the Data nowadays is cloud-based. No human has nearly as much memories as the clouds have.
Especially imagine advertisement, where the creative has to come up with ideas that resonate with their target group. An AI can analyze its target group very, very, very good. So naturally, it can make very, very, very good decisions based on that knowledge. So what color should the background be? The color that most people resonate with. What font should it use ? The font that most people resonate with. Whatever those decision are based one, the result will be effective.
I think the main difference between AI and the human creativity is this phenomenon of breaking out. In german we say "to look over the edge of the plate", basically breaking out of your own comforted thoughts and thinking of weird and random things. That is what makes alot of artists stand out and what makes art special compared to just plain advertisement.
But also this could be achieved by AI and this behaviour of randomness and arbitrariness is often used software dev. especially in game development, for example when programming Navigation AI that sometimes chooses "humanly looking" actions over the best and most straightforward ones.
I agree with you though, that AI does not have to be feared in the creative industry. As many have said before, adapting to change and advancement is a natural process. This has happened many, many times before to every economic field there ever was and ever will be.
From Cambridge Dictionary, Imagination:
- The ability to form pictures in the mind. Something that you think exists or is true, although in fact it is not real or true.
- The ability to think of new ideas.
A machine can't think in new ideas, it just executes an order/command/instruction of mixing ideas previously defined. We can't expect much more. In contrast to the machine, we may form new ideas in our mind at will, and more, without using our experiencie taken from Reality (there's a spanish saying: "even a blind can imagine").
The simpliest way of mixing ideas is using all the info we get through our eyes or vision, the visual info (we also can get spatial info). But that alone doesn't imply we have developed a good or true imagination. The machine needs not only our vital experience, but the vital experience of millions of humans to achieve something that may be "plastic" artistically talking, not believable (we see instantly: "that's not real", "not new", "not cool", "not beauty"). Using big data, an AI could choose which color is indeed better for a product -chosing the most liked one, easy-, but that's not what i was talking about. There's no app, nor AI that can make decissions for me. And choosing the most liked thing in instagram for example, is not something to be proud of, it's very simple.
Sadly, in an artistic way, a machine would mix things like greeks did a long time ago, like mixing eagle+lion=griffin. And that, in my personal opinion it's not imagination, nor something creative, just childish, ridiculous and plain stupid. Furries included.
Like i said with other words, we also can imagine things we never saw, and that's a fact nobody can negate. We can create forms that never existed in Reality, just using our sense of abstraction, or dream "we are flying". Not all people have that ability developed, the ability to create new things from nothing, "like from the void". And when we see "something new", we tend to say things like: "wow, this guy is incredible, it has true imagination, i could never imagine that!".
In our mind we can play with "atoms", -concepts out of reality-. Could a machine "play" like us? i'd say that Never (a machine has no motivation). The more complex the mixing, the worse for the machine because it doesn't understand human logic at all, just a mathematical logic we programmed on it (AND, NOT, OR, TRUE, FALSE, etc.). Could a machine create a car? a new form of vehicle that actual works? a car design? a life form using genetics and not falling in chimeras? i doubt it. And you know why, because an AI, a machine lacks a very important thing: WILL and ability to understand things.
Nowadays, the most advanced AI just recognizes things with a failure rate pretty high. There was an accident with a Tesla car that didn't recognize what it was a person, and that person is actually dead. It's not very logic to give AI too many expectations.
A creative example: if you "create a new soldier character", with camo variations, helmet variation or weapon variation, do you really think you are creating something new? the response is no, you don't. You would be just mixing, mixing something like images of "lego bricks", bricks that other person created. Other thing is to create the lego bricks and all what it means, and for that, we need imagination. A machine/AI could never create something like LEGO. This is what makes the difference with artists, some people use Daz Studio girls/basemeshes/autoskin plugins, and others not. Procedural tools won't make us better.
I think the true nature of creation is behind all that. Taking as example the "soldier's weapon" or "legos", did you ask yourself why a weapon has that form? why that design?, or why its parts are made in the way the have been made, or did you know the history of weapons and its evolution?.
I think that as "artists", we opt for the easiest way of "creation", we tend to mix things a lot, we tend to use the same kitbashes, basemeshes, libraries, and we don't understand too many things. It's one of the reasons of why almost all artist need to study human anatomy before making a proper human. We may think AI will solve all our artistic problems, and that's not true, it won't happen, never.
I'm not going to convince you, i know. We all know our brains have two parts, one rational and logic, and irrational. Logically speaking, i have zero expectations with AI, and more watching like AI in gaming doesn't evolved anything in 40 years. So i don't want to fall into irrational expectations about the future of AI. Our technology is not infinite, and we have limits for all. Too many films failed in an epic way with its "predictions" and we are like stucked with our technology, more than a decade. I'm still waiting for a 100TB HDD.
The day we understand our mind, how it really works, and too many other things, then, we might make artificial humans or new life forms using maybe 3d printers?. But being realistic, we are not gods, and that's dreaming hard
There is typically little room for "playing around" and coming up with things on the fly. Usually you are not hired because you are an "artist". You are hired because you can get a concept or an image and turn it into a prop or whatever is needed.
And depending on the game imagination has a small role..if you are making a trashcan, a tree or a car, you don't need much imagination, just some basic guidelines and something that fits the style of the location it will be placed in. And style transfer is something AI is getting good at.
"Nowadays, the most advanced AI just recognizes things with a failure rate pretty high. There was an accident with a Tesla car that didn't recognize what it was a person, and that person is actually dead."
-Self driving AI is statistically safer than human drivers..not sure if that's a fair comparison at this point, as I think AI is mostly tested on cruise mode and not in complex city driving, but humans makes dumb ass mistakes all the time. Not sure why you'd judge the AI more harshly.
And saying the failure rate for recognition is high isn't the whole truth. It's worse at some things, better than humans at others.
I think you are right in some limitations that might be hard to get past. A human can for example have "happy accidents" and think hmm that looks pretty cool and go with it...I'm sure an AI has tons of "accidents" but I'm not sure how it would identify them as good or not.
You could also use human guided AI for the best of both worlds.
I think the potential for technology is virtually unlimited. As in if it's theoretically possible, it will be possible at some point, when we have full control of the basic components of matter.
How are atoms out of reality ? I think Einstein had a solid proof for atoms existence
Can AI design a new form of vehicle?
Take a look at this AI designed car:
https://www.fastcompany.com/3054028/inside-the-hack-rod-the-worlds-first-ai-designed-car
Of course this design is based on logic, although there are concepts of logic that go beyond the classic ones you listed.
Fuzzy logic for example:
https://en.wikipedia.org/wiki/Fuzzy_logic
When humans designed cars they did so, by following logic. They asked questions: How do we get faster from A to B ? Then they came up with a solution. Thats basically what an AI would do to solve any problem.
"a machine lacks a very important thing: WILL and ability to understand things. "
No, this ability to understand and learn things is the basic concept of modern AI.
https://en.wikipedia.org/wiki/Machine_learning
I dont want to convince you of anything, but I think you are plain wrong in assuming that high expectations on AI are not very logic.
Of course they are not as developed as other technologies, yet. I am sure if you read some about this new technology, you will change your mind on your own. AI is the future. If we will witness that future is a different question, but it is very likely for AI to advance and offer solutions for so many use cases including solutions where imagination and creativity is required.