People seem to think generative tech is just going to progress faster and faster. This completely disregards that the costs to noticably improve image generators or LLMs is increasing exponentially. It's not economically feasible to make better and better models if the next model costs magnitudes more than the last. Available training data isn't unlimited either. (Btw 3d training data is soo incredibly scarce in comparison to image and video data, if anyone is safe, it's us) I think the stagnation is already becoming visible. This tech won't replace us any time soon.
@@NycroLP Sure 3D generators do improve, but there's a cap to the progress of this tech because at some point the return of investment can't keep up with the exponentially increasing costs of bigger models. And because the resource for the progress is training data wich is limited.
I agree completely. I've heard and seen so many people completely misunderstand how diffusion image generators work, misinterpret the controversial points, and (just as they did when CGI first began to be used in visual effects) overestimate both the ease and the results. You did a great job explaining how these systems work, and why they often don't. One thing I wanted to point out: Adobe Firefly source material isn't as cut-and-dry as they tried to make it sound. People were shocked to discover that their work was being used to train the A.I. model simply because they'd signed up for and used an Adobe service that existed before this technology. They were saying they never agreed to it; Adobe had this material of theirs in their vaults, as it were, and decided that it being there made it fair game for them to use.
We’re implementing Ai in our architecture firm and it can create interesting designs. However, We don’t have a lot of control over what it produces and what it often designs are things that are physically impossible to build.
Thanks for a very fair and clear summary. One thing I would add is the very basic problem of using words to define images- to me this represents a fundamental limitation on how controllable these systems can be. The old saying that 'A picture paints a thousand words' is a recognition that the visual has a complexity and a nuance that cannot really be expressed with words. So at the very root of this technology is a misunderstanding, which is the notion that by using word combinations I can fully describe the image I want the system to make, and thus have full control over the output. This is not true- no amount of words could really fully describe a persons face, for example, and anyone who doubts this should try generating a self portrait using AI by typing in a description of their face- it just can't be done.
While I agree with all your technical observations, I think the outcome will be very different and I already see it happening in advertising and ecommerce, simply because you apply your own standards (technical to ethical) and dont consider that people in the position to decide are often not concerned with either quality or ethics at all - they will gladly run AI movie posters or ads and will laugh in your face when you mention trademarks, copyright, consistency or output quality - its totally worth it for them to compromise while saving a couple thousand dollars/Euros. This is the clients dream: finally they have all the petty creative decisions in their hand without having to pay more than 12 bucks and do everything in-house. The are already AI ads in the streets and all over the internet. Will big film studios use AI in wfx the way we know it? Probably not right now. How many people work in big VFX houses though, perhaps a couple thousand? 99 % of 3D artists will be somehow affected, just like photographers, illustrators and others.
@@NeonLuminous How often do you run ads with Disney characters recently? Thats not the case in the slightest. The idea that AI will not be implemented on a massive scale just because "you are potentially responsible for trademark infringement in case of over-fitting" is just laughable, just like saying that clients wont ever use stock photos because they only own editorial license and cannot print it out on billboards - they simple DONT CARE, 1 in a million might get caught, the rest will gladly take a shot at it to save money.
I say this all the time when I go to AI generator videos and I ask the people there, what is this technology actually good for? Other than the tiny amount of personal enjoyment you might get out of generating 4,000 images of Iron Man. I just can't understand all the hundreds of millions maybe even billions of dollars that are being invested in this technology, when nobody seems to have any clear goal for what it's good for? Right now AI art seem to fill no niche and serve no needed purpose. And yet in spite of that we're pumping out millions of images every single day that no one's ever going to see and that serve no purpose of any kind
Copyright is dead. That’s THE big issue about AI. And yeah, it will replace sweat shop workers at all the big studios. So it’s liberating actually. But what will those people that never complained do, once they are set free? I guess working at Costco, wearing Crocks. Edit: concerning Adobe, are you living under a rock? They trained their AI on personal files on Behance and from “your” cloud on Adobe servers in general. Plus every time one uses the remove tool, or generative fill, the image on the screen get’s transmitted to Adobe and is used to train the AI.
I completely agree on all points. This is a great summary of the state of AI image generation and the issues with it. These are exactly the same problems I've encountered while experimenting with it. Ultimately, as it stands currently on Aug 28th, 2024, it's pretty useless for anything related to professional artwork pipelines. It's hard to predict how much potential there is for improvement and how long it will take. But the problem of hallucinations, understanding form, consistency, iteration, etc., do seem like they'll take some time to solve, requiring further breakthroughs in technology. And on top of that, it needs an astronomical amount of reference data to train on, which is impossible to acquire without infringing on copyright.
"it needs an astronomical amount of reference data to train on, which is impossible to acquire without infringing on copyright" AI companies already did that, and will continue to do that until there are regulations in place to stop it. Unfortunately the rate at which this technology is being developed outpaces any means of regulation to begin with.
Adobe recently got in big trouble by stating that anything uploaded to their cloud storage would be used to train Firefly. This created a frenzy of professionals scrambling to remove copyrighted work from the cloud. Remember, a lot of professionals work on things they do not own the copyrights or trademark to. And then adobe got in trouble for making it extremely difficult to cancel subscriptions to their software. So Adobe isn't really the way to go. They were exploiting artists long before AI, and now they are just getting ridiculous about it.
Another thing I’ve heard is that everyone can use AI, but only professional artists know how to get a good result. Amateurs will generate something. Within a few prompts, they could get something realistic and decide it looks good. However, a professional artist has the eye to properly critique it and decide what specifically needs changed to make it look better.
@@reaganmonkey8 that's a good point. It doesnt matter how technically realistic an image is if it's supposed to convey one feeling but the camera angle and lighting clash with another feeling
People and corporations see art as just another commodity, a product. Something to slap onto an advertisement or to sell as is. How it gets made isnt any of their concern. What is their concern is however the time, money and obligations they have to devote dealing with humans that produce art. Remove the human and many of the things they hate about art vanish and they are free to exploit every piece of media they see fit. This is why their pushing to replace you, because they hate you and the fact that they are forced to involve you in their business to make profit.
I agree, it's so nice to hear someone talk about it level headed. I think , even if all the technical issues were 'solved', it wouldn't be so much of a threat because, at the end of the day, people look at art because it's the expression of a human artist. people just don't want to pay money to see what a computer draws no matter how good it is, but as a tool to occasionally help artists the way you've described, it's useful. I'm just not sure it's useful enough to ever make a profit.
Hm, well, I agree to this about videography. AI not having an impact on artists, I would disagree. There are a few areas where this is already having a severe impact. For example one I wanted to work in - product photography. Not really for the product itself, but the background, especially for anything social media. To a degree you could also argue that video backgrounds could already be generated for that. Want to quickly make the bottle you just shot appear in a jungle or some arbitrary nature scene? Not a problem, especially not if you ask it to generate it with a lot of bokeh. Also all the AI-based infill methods in Photoshop are above and beyond other tools in other software. And AI is really good with color schemes and general composition/contrast. I DO think that AI will make a lot of people jobless the same way as smartphones and smartphone cameras also killed a lot of the "regular" photography jobs and also the compact camera market. It's not about how good you can make a product shot or a group photo of kids at school. It's about how good is good enough, how easily that "good enough" is achievable and how much cheaper it is for the customer. And it's also about efficiency. If AI is boosting my and everybody else's efficiency by say 20%... I would suddenly need 20% more jobs. Or 20% of people in the field are simply not needed anymore. Jobs being made obsolete by new technologies isn't a new concept. It happened to film cameras. It will to a degree now happen to digital photography and with likely a bigger delay also to videography. Though that one is much harder. But anything still image... or writing... it will be rougher.
Cave painters were replaced by pottery painters who were replaced by hieroglyph painters who were replaced by mural painters who were replaced by wood panel painters who were replaced by canvas painters who were replaced by digital painters who were replaced by AI which was replaced by cave painters after it all broke down. Then one day, cave painters were replaced by pottery painters ...
Good observations. Pretty much same things that I noticed. Tried to use it for concepts but it's useless because it can't draw logical details - just spews nonsense.
I think Ai voices are taking over really quick. You're seeing all kinds of creators even making ai versions of themselves just to make production easier. I also think that AI will have a larger effect on the video game industry. Which might be good because some of these studios are massive and no longer focus on gameplay.
One thing people forget is that we get "commercial" tool. We do not really know what is behind curtain. Remember it's new tech raise, either you're first or you'll left behind. There is huge money in it and I have huge doubt what tools we are given is what giants really have. As stated many times, we are in times of remakes and reused poop, either gaming or movies, vast majority of new releases are not original. If "people in suits" can save money, they will. And they could do it with AI, in future
I would agree with this take, like I did with Feng Zhu when I watched his video, especially about problem solving and technical details, with me coming from a technical/trade background. I can see it being useful for quick idea tossing in the beginning of a project or for some background stuff as you said. Personally I want it, and I bet studios would support that, to take over the tedious and time consuming stuff like rotoscoping, UV unwrapping and so on. That way artists would have more time to focus on the creative parts and studios could save money, I assume. Eventually it probably will get better and then it's up to us artists to make use of the tool (that's what it is, just another tool) and integrate it to improve the workflow. However all the panic is exaggerated in my opinion. From a different perspective I've seen this many many moons ago when CNC machines became feasible in the cabinetry industry and everybody saw doomsday coming and cabinetmakers lose their profession. And yet here we are and those who implemented the tool thrived. Even those that didn't are still kicking and doing well.
Since AI is just a tool and not true AI then it isn't likely to replace artists, but it will probably become a popular tool used by non-artists to circumvent employing artists.
I agree that it won't cause Massive Industry change, but it in terms of anything other than the professional level, AI will take over. It produces 'good enough' images and video that will drown the internet in scum. It's capacity to produce mediocre images at such quanitity is still worrying.
Just a plug about actors looking different throughout the movie... ever seen T.Gilliam's "The imaginarium if doctor Parnasssus"? ;). Great video btw! Lots of solid points.
Something AI misses is how specific you, yourself, can get with your own art. You can perfectly design everything you want in any way to the pixel level. Even with stuff like controlnet, loras, everything, you can't change it down to the pixel.
All of the critiques like this seem to ignore how new this technology is. If you're saying "AI can't do this" you really need to tack "yet" onto the end of the sentence
Welp if yer a kinda guy who says it will replace Artists- extrapolate your carelesness up untill it reaches your forhead and you get that if you let this happen once it will happen to you sooner or later No empathy for someone's job means noone will empathise will your job once you are in danger
If AI gets advanced enough, I think our jobs might change from doing most of the process manually like we do now, to us become a type of lead 3D artist overlooking a team of AI. For example, we might use AI to generate a model, retopologize and generate a texture. Then we would step in to polish, fix and change what the AI outputs to make sure it's consistent with the project's style, and technical parameters. Even though image generators can make artworks that look great, entrepreneurs and producers still hire artists to overlook the creative process since our expertise is not just knowing 3D, but also being able to take the right creative choices for the job. I think that 3D artists not using AI in their workflow might fall behind 3D artists using it.
Rob I've seen so many explanations of diffusion models but realized I didn't truly understand it until your explanation. You have a real talent for breaking down complex topics!
@@DECODEDVFX😮then your doom to not be able to live in the future😮 you never thought of it so it will be incomprehensible to you how is it possible I knew AI would be here around now I know exactly where we're going😮 or much more than you do because I'm paying attention you are not😮 you don't know about brain organoid computers😮 you don't know that right now a guy is being used to map all the neurons in his brain with AI😮 you don't know a new AI model is coming out that has 300 million contacts length😮 I do I can see and predict the future the future is happening right now😮 you're literally just talking about technology from a year ago and complaining about it😮 you're doomed your whole your people you're doomed😮 you should be telling people AI will take their jobs and they need to prepare for it you are literally making people starve in the street you in a way you're killing them😮 you're killing them with the LIE of Hope😮 multi-billion-dollar multi-trillion dollar corporations they don't give a fuk about your copyright your your artist integrity if AI makes more money you're doomed and AI does make more money😮 AI will continue to make more money and if they need to make a nuclear power plants and you and you have to live in the dark with no electricity to power the AI so be it that's what they'll do😮 you think because you find it morally unacceptable that it's not going to happen😮 what are you people children
There is no future in which AI replaces artists, unless that future is AI killing all humans. AI is not going around smacking the paint brushes out of people's hands. Even if all work is automated and we have nothing to do all day someone will be making art, in fact I would bet more people would be making art. It's one of those weird things humans just seem to do, like its baked into our nature. Will it replace some (or many) forms of commercial art? Yeah probably, but I see that as a good thing. Work culture in the US is in severe need of reform and the arts are the best place to start having that conversation.
Well this is what many people have said: that art will return to being a rich people’s hobby. Most people simply won’t have enough leisure time because they will likely have to work multiple jobs to pay their bills.
Similar issues exist in the world of software development. All AI generated code does is save me the effort of wading and sifting through scores of stackexchange posts. I still have be able to look at it produces and determine if it is correct. Even if it is correct, it might still not be usable directly. Will AI negatively impact jobs? Yep. There are people who will lose their jobs. Is everyone going to lose their jobs? No. Skills, thinking and competencies are still needed. AI is just another tool.
Yeah. I'm not much of a coder. I used chatgpt to help me with python stuff. The code is quite often nonsensical and when it works, it's often bloated and inefficient. It provides a decent starting point for a novice like me, but it can't replace clean code written by professionals.
@@DECODEDVFX AI often is very good at giving you either a basic structure for something or sometimes also good at outlining bugs or debugging things, especially when you are mentally stuck on one angle. I used ChatGPT recently to generate a UI component in a framework I wasn't really familiar with which I just had to rework and read up on it - that likely saved me about 50% of the time of having to figure it out from scratch myself. That's an extreme example I would say, but it does boost efficiency and the increased throughput in workload means we either need more work or people will lose their jobs.
Coding is a really interesting case, since according to a survey, the majority of programmers enjoyed using ChatGPT in their workflow, however at the same time programmers using AI tended to have worse code. Like ChatGPT works great if I just need to write an simple addon for Blender, but I bet it falls apart when developing something more complex like flip fluids
AI Music is scary good, but it doesn't really give a crap what you put into it if you provide it a sample, which makes it feel like it's just crossfading someone's existing work, not easy to tell when that happens though.
Well I wouldn't call it good music, it's great at churning out tons of generic garbage the elevator music type stuff but I have yet to hear an AI song that had any real heart or that moved me in any way.
@@aegisgfx It sounds like something, but kind of like image AI sucks at making hands and feet, AI music also has some weird issues that get in the way of writing something that'll move you I will admit that. But music does seem like an area AI does a little better in - Unless it's just a hoax for freelance producers who take peoples requests and just slap something together in minutes, idk lol
The thing all seem to be missing about the AI debate (seen many on RUclips chime in on this) is the historical context. The conversation currently going on about AI has been made many times before, every time tech and innovation affects art. One clear example is the camera. Photography was even called by the famous poet Baudelaire "the refuge of every would-be painter, every painter too ill-endowed or too lazy to complete his studies". Basically the same quote is said now about AI. Just a few years ago, the same was said about people using photobashing for concept art. This is just the same debate going on over and over, and I barely see anyone on RUclips mentioning it. This could also be said about the whole "AI is not art"-statement some people claim. The discussion about what is "art" has also been going on for a long time, but still no one brings up important works like Duchamp's Fountain in these types of videos. P.S. Not necessarily claiming the video argued all of the points brought up above, just saying history is a nuance which is missing in the general AI debate.
From what I've seen, this historical context is exactly what AI proponents often bring up at one point or another. I would like point out that the camera, while absolutely as revolutionary and upending to the art world at the time as AI is now, was developed in isolation of artworks (although it more or less built off of technology that artists had used for realistic renderings previously). The same can be said about digital art software, which offers an alternative -- and often more efficient -- approach to traditional processes. Generative AI for images and other forms of media, on the other hand, could not function without the millions of artists and billions of artworks that (unknowingly) stand behind it. The generation of work from a simple text prompt is completely devoid of process, and is more of an exercise in vocabulary and syntax that is ultimately guaranteed to return good results than it is a tool. Tldr, generative AI is not a means to an end like photography or digital art programs. It IS the end. It does all of the work for you, and it could not function in absence of reference material.
@@OctaveIndustries Your comment unfortunately makes little sense to me. Photography is the press of a button in its simplest form. How is that process more advanced than writing a prompt? The core difference is that photography now is accepted as an art form (but it was not during the time of Baudelaire). Photography, in its simpler form, is a lot about having the eye - finding interesting shots and capture something. Same about AI. To make interesting art with AI you need the eye, to direct the AI into making something valuable. And if directing is not an art form, is a film director not an artist? The film director does not do the photography, or the acting or the writing (usually that is). He/she directs. And what about a VFX supervisor? The issue with the anti-AI movement is that it doesn't make sense if you judge AI art the same as previous art.
@@OctaveIndustries exactly. people can claim that coming up with prompts is 'just as hard and requires skills', but at the end of the day, the image they generate isn't created by them, but by a machine. Imagine giving an idea to a painter to draw, and after seeing the end result claiming that they were the ones who painted it. It's the exact same thing.
Totally agree that people who want to do VFX should do it. Disagree that AI "cannot do things" based on current state.
2 месяца назад+1
Also, I don't think you have mentioned that you can use AI to "improve" sketches (either be line art, or more developed color drawings). One more thing. What's the difference between AI learning from other artists or a person doing exactly the same thing?
Okay, start learning from the works of Rembrandt, or Vereshchagin, if you don't want to live and want CGI, start copying wlop or any other artist. I'm sure your result will be different. And AI will identically steal pieces of work, but in fact this is not intelligence at all, it is marketing that made it intelligence. + Using the works of artists when learning AI is illegal because you are literally using someone else's work to make a profit. I am an architect, I am happy with the topic of AI development because it simplifies routine tasks, but it cannot sign papers and bear responsibility) But I am very afraid of the issue of devaluation of art, because it can lead to stagnation. Art is what from early times distinguished a thinking person from a monkey. I hope the value of human art will not fall in high circles, but will only increase, and that the development of AI will not lead to the fact that most people will become like monkeys.
2 месяца назад
@@denkot442 Why do you say it is not intelligence? How are you different from an AI?
It's seems there's a significant knowledge gap in what you think AI does and what it now can actually do. With multimodal AI almost all of these issues are resolved/going to be resolved in the next 6-12months. Besides, I've been thinking about the differences between AI and humans, has there ever been good artist who never in their life copied someone else in order to learn?
@@DECODEDVFX It reminds of that saying "Great artists copy, bad artists steal". I don't think it's about copying, but being able to take inspiration and using it in a new way. Since it's almost impossible to make something completely new. However if I used AI to generate a picture then post it on Instagram and saying it's made by me, that would be wrong. However then again, the client does not care about that, they just want good results
I don't want to be that guy buuuuut... a lot of the flaws and weakenesses you mentioned in here are being worked on right now. I'm not an AI artist but i've been dabbling with it for some time and people are finding very creative ways to work around those flaws and getting better at using AI
2 месяца назад+1
I think that you're right talking about AI now; but this is like comparing Blender today with the first 3D program. AI for art has just started (less than 2 year ago, I believe). We are just at the very beginning. Also, I don't think you can say that AI doesn't understand "form". Clearly it does, and it is getting better. To me it is like looking a drawing by a 5 years old and saying that humans don't understand form.
Yes, AI will replace specialised artisans completely. But it will offer everyone the tools to make their own creations. If you want to watch a movie with Charlie Chaplain, Brad Pitt and Elon Musk in it... in a musical remake of Casablanca with a happy ending. AI will make it instantaneously for you to enjoy,. Or in 3D, in a virtual headset... with you as the main protagonist. But the most important aspect, is that AI will teach itself. It will create new forms of entertainment and 'art' It will free people from the laborious task of making stuff... to being the producers of your own entertainment. Remember, with AI...your watch ( if you can afford it ) will be your doctor. A career as a doctor...will be over. The AI enabled future will revolutionise everything.
@@SMCwasTaken Landscape artists are hobbyists. Just like knitting... it's a pastime, a repetitive task in order to disappear from the world and yourself.
I just can't get enough of the fact that nobody cares about stock authors and for some reason using stock database is justified. Why so? becouse authors agreed to these? It;s true, but nobody asked them beforehand! They just got it as ultimatum. More so, once you agreed, you can't opt out. They will have all your images, photos and illustrations embedded in AI model, forever! Oh, and you also will not get any of payment, so.
Yup. Look on game developers, most of the giants had been releasing rehashed and reused poop for years now. If they gonna get things done quicker and save money they gonna do it and use AI.
you basically described, how humans learn. Children go to school, teachers show them renaissance pictures and tell them this is a renaissance picture until they can recognize them quite well. That's also how you learn new skills. first you might make many mistakes, but you train again and again and later you will use your skills very well. And retraining is difficult for humans too, once you learned something and used it for years, it's hard to use something new. Just saying... And as concerns to copyright: people may know the Mona Lisa and then they start to draw these images themselves. Are they violating copyright laws? According to your logic yes. Basically as soon as you see any image / painting, you need to buy a license. right?
You have a handle on it? You know what it can do? Those are some bold statements. Do you know what it’s going to be able to do next week? Because That’s what people are worried about. Not this week but next week and the week after. And next year!
The problems with the current AI models are inherent to how diffusion models work. They can be improved over time, perhaps, but they can't be fixed. Future models may solve these issues, but right now, this is the best they can do.
Nope. Ai doesn't have life experience and feelings, just knowledge. He's like an academician, not a creator/artist. And human pretending being artist using Ai has just prompts to proove his productions. Real artists have original éléments, sketches, roughs, researches etc.
@@comebackguy8892 probably ;) but my tone was drastic. I am not criticizing those who only use AI. These people may have fantastic ideas but not the talent to bring them to life. And conversely, a talented artist is not immune to running out of inspiration.
Definitely! AI will make it easier for us to do the projects we wanted. Even though anyone can generate an AI image, artists will still be needed to make the results actually good. It's like everyone can take photos with their phone, but we still admire great photographers
Everyone talks about how AI is stealing work. But artists also “steal” other people’s work for inspiration. An AI can have so many pictures that it takes from all of them and isn’t necessarily copying one. There are exceptions, like that girl he mentioned, but I think it would be fixed with more data.
I want AI to do my housework, not my creativity.
People seem to think generative tech is just going to progress faster and faster.
This completely disregards that the costs to noticably improve image generators or LLMs is increasing exponentially. It's not economically feasible to make better and better models if the next model costs magnitudes more than the last.
Available training data isn't unlimited either. (Btw 3d training data is soo incredibly scarce in comparison to image and video data, if anyone is safe, it's us)
I think the stagnation is already becoming visible.
This tech won't replace us any time soon.
AI/ML doesn’t have to exceed humans to displace them in the workplace.
Are you sure about that? 3D generators seem to get noitceably better every few months.
@@NycroLP Sure 3D generators do improve, but there's a cap to the progress of this tech because at some point the return of investment can't keep up with the exponentially increasing costs of bigger models. And because the resource for the progress is training data wich is limited.
Because they are new. This sort of tech always improves in massive leaps at the start.
@@eye776 True. But i don't believe it will have as much of an impact as many fear
In conncept art, especially characters, every little line and curve changes the look of the character entirely. Ai just cant handle that.
I agree completely. I've heard and seen so many people completely misunderstand how diffusion image generators work, misinterpret the controversial points, and (just as they did when CGI first began to be used in visual effects) overestimate both the ease and the results. You did a great job explaining how these systems work, and why they often don't. One thing I wanted to point out: Adobe Firefly source material isn't as cut-and-dry as they tried to make it sound. People were shocked to discover that their work was being used to train the A.I. model simply because they'd signed up for and used an Adobe service that existed before this technology. They were saying they never agreed to it; Adobe had this material of theirs in their vaults, as it were, and decided that it being there made it fair game for them to use.
You both understand how AI works on every level. You studied it probably. i would give you 2 a nobel price in machine learning.
AI will never use a small can of WD40 to loosen the lid of a large can of WD40, and as such it will never truly be human.
Neither will most humans
@@Therealnonsequitur They are probably cursed butane-barbecue users
Best comment!
😂😂😂
We’re implementing Ai in our architecture firm and it can create interesting designs. However, We don’t have a lot of control over what it produces and what it often designs are things that are physically impossible to build.
Thanks for a very fair and clear summary. One thing I would add is the very basic problem of using words to define images- to me this represents a fundamental limitation on how controllable these systems can be. The old saying that 'A picture paints a thousand words' is a recognition that the visual has a complexity and a nuance that cannot really be expressed with words.
So at the very root of this technology is a misunderstanding, which is the notion that by using word combinations I can fully describe the image I want the system to make, and thus have full control over the output. This is not true- no amount of words could really fully describe a persons face, for example, and anyone who doubts this should try generating a self portrait using AI by typing in a description of their face- it just can't be done.
While I agree with all your technical observations, I think the outcome will be very different and I already see it happening in advertising and ecommerce, simply because you apply your own standards (technical to ethical) and dont consider that people in the position to decide are often not concerned with either quality or ethics at all - they will gladly run AI movie posters or ads and will laugh in your face when you mention trademarks, copyright, consistency or output quality - its totally worth it for them to compromise while saving a couple thousand dollars/Euros. This is the clients dream: finally they have all the petty creative decisions in their hand without having to pay more than 12 bucks and do everything in-house. The are already AI ads in the streets and all over the internet. Will big film studios use AI in wfx the way we know it? Probably not right now. How many people work in big VFX houses though, perhaps a couple thousand? 99 % of 3D artists will be somehow affected, just like photographers, illustrators and others.
Me laughing at Disney's face while they send me a cease and desist letter. 🤣
@@NeonLuminous How often do you run ads with Disney characters recently? Thats not the case in the slightest. The idea that AI will not be implemented on a massive scale just because "you are potentially responsible for trademark infringement in case of over-fitting" is just laughable, just like saying that clients wont ever use stock photos because they only own editorial license and cannot print it out on billboards - they simple DONT CARE, 1 in a million might get caught, the rest will gladly take a shot at it to save money.
I say this all the time when I go to AI generator videos and I ask the people there, what is this technology actually good for? Other than the tiny amount of personal enjoyment you might get out of generating 4,000 images of Iron Man. I just can't understand all the hundreds of millions maybe even billions of dollars that are being invested in this technology, when nobody seems to have any clear goal for what it's good for? Right now AI art seem to fill no niche and serve no needed purpose. And yet in spite of that we're pumping out millions of images every single day that no one's ever going to see and that serve no purpose of any kind
It's not about the art. Its about understanding the world.
Highly under rated video! Preach it, man 🗣️
Copyright is dead. That’s THE big issue about AI. And yeah, it will replace sweat shop workers at all the big studios. So it’s liberating actually. But what will those people that never complained do, once they are set free? I guess working at Costco, wearing Crocks.
Edit: concerning Adobe, are you living under a rock? They trained their AI on personal files on Behance and from “your” cloud on Adobe servers in general. Plus every time one uses the remove tool, or generative fill, the image on the screen get’s transmitted to Adobe and is used to train the AI.
Doubt it will end copywrite.. more like copywrite will keep AI at bay
I completely agree on all points. This is a great summary of the state of AI image generation and the issues with it. These are exactly the same problems I've encountered while experimenting with it. Ultimately, as it stands currently on Aug 28th, 2024, it's pretty useless for anything related to professional artwork pipelines. It's hard to predict how much potential there is for improvement and how long it will take. But the problem of hallucinations, understanding form, consistency, iteration, etc., do seem like they'll take some time to solve, requiring further breakthroughs in technology. And on top of that, it needs an astronomical amount of reference data to train on, which is impossible to acquire without infringing on copyright.
"it needs an astronomical amount of reference data to train on, which is impossible to acquire without infringing on copyright"
AI companies already did that, and will continue to do that until there are regulations in place to stop it. Unfortunately the rate at which this technology is being developed outpaces any means of regulation to begin with.
Adobe recently got in big trouble by stating that anything uploaded to their cloud storage would be used to train Firefly. This created a frenzy of professionals scrambling to remove copyrighted work from the cloud. Remember, a lot of professionals work on things they do not own the copyrights or trademark to. And then adobe got in trouble for making it extremely difficult to cancel subscriptions to their software.
So Adobe isn't really the way to go. They were exploiting artists long before AI, and now they are just getting ridiculous about it.
People who work in that industry who have not abandoned Adobe yet, I really have no sympathy for people that dumb.
Another thing I’ve heard is that everyone can use AI, but only professional artists know how to get a good result.
Amateurs will generate something. Within a few prompts, they could get something realistic and decide it looks good.
However, a professional artist has the eye to properly critique it and decide what specifically needs changed to make it look better.
@@reaganmonkey8 that's a good point. It doesnt matter how technically realistic an image is if it's supposed to convey one feeling but the camera angle and lighting clash with another feeling
People and corporations see art as just another commodity, a product. Something to slap onto an advertisement or to sell as is. How it gets made isnt any of their concern. What is their concern is however the time, money and obligations they have to devote dealing with humans that produce art. Remove the human and many of the things they hate about art vanish and they are free to exploit every piece of media they see fit. This is why their pushing to replace you, because they hate you and the fact that they are forced to involve you in their business to make profit.
I agree, it's so nice to hear someone talk about it level headed. I think , even if all the technical issues were 'solved', it wouldn't be so much of a threat because, at the end of the day, people look at art because it's the expression of a human artist. people just don't want to pay money to see what a computer draws no matter how good it is, but as a tool to occasionally help artists the way you've described, it's useful. I'm just not sure it's useful enough to ever make a profit.
Hm, well, I agree to this about videography. AI not having an impact on artists, I would disagree. There are a few areas where this is already having a severe impact. For example one I wanted to work in - product photography. Not really for the product itself, but the background, especially for anything social media. To a degree you could also argue that video backgrounds could already be generated for that. Want to quickly make the bottle you just shot appear in a jungle or some arbitrary nature scene? Not a problem, especially not if you ask it to generate it with a lot of bokeh.
Also all the AI-based infill methods in Photoshop are above and beyond other tools in other software.
And AI is really good with color schemes and general composition/contrast.
I DO think that AI will make a lot of people jobless the same way as smartphones and smartphone cameras also killed a lot of the "regular" photography jobs and also the compact camera market. It's not about how good you can make a product shot or a group photo of kids at school. It's about how good is good enough, how easily that "good enough" is achievable and how much cheaper it is for the customer. And it's also about efficiency. If AI is boosting my and everybody else's efficiency by say 20%... I would suddenly need 20% more jobs. Or 20% of people in the field are simply not needed anymore. Jobs being made obsolete by new technologies isn't a new concept. It happened to film cameras. It will to a degree now happen to digital photography and with likely a bigger delay also to videography. Though that one is much harder. But anything still image... or writing... it will be rougher.
just waiting when this hype of AI crab ends. I hate it is forced to hardware itself (to increase price).
Cave painters were replaced by pottery painters who were replaced by hieroglyph painters who were replaced by mural painters who were replaced by wood panel painters who were replaced by canvas painters who were replaced by digital painters who were replaced by AI which was replaced by cave painters after it all broke down. Then one day, cave painters were replaced by pottery painters ...
Good observations.
Pretty much same things that I noticed.
Tried to use it for concepts but it's useless because it can't draw logical details - just spews nonsense.
There's also AI cleanup and upscaling, but my workflow doesn't allow for much.
I think Ai voices are taking over really quick. You're seeing all kinds of creators even making ai versions of themselves just to make production easier. I also think that AI will have a larger effect on the video game industry. Which might be good because some of these studios are massive and no longer focus on gameplay.
Sound is easier to modify.
Bigger threat atm than Ai is outsourcing to other countries for cheap labor
One thing people forget is that we get "commercial" tool. We do not really know what is behind curtain. Remember it's new tech raise, either you're first or you'll left behind. There is huge money in it and I have huge doubt what tools we are given is what giants really have.
As stated many times, we are in times of remakes and reused poop, either gaming or movies, vast majority of new releases are not original.
If "people in suits" can save money, they will. And they could do it with AI, in future
Race... not raise...
Are u the sort of person who doesn't change their mind no matter how much logic is thrown at them.
I would agree with this take, like I did with Feng Zhu when I watched his video, especially about problem solving and technical details, with me coming from a technical/trade background. I can see it being useful for quick idea tossing in the beginning of a project or for some background stuff as you said. Personally I want it, and I bet studios would support that, to take over the tedious and time consuming stuff like rotoscoping, UV unwrapping and so on. That way artists would have more time to focus on the creative parts and studios could save money, I assume.
Eventually it probably will get better and then it's up to us artists to make use of the tool (that's what it is, just another tool) and integrate it to improve the workflow. However all the panic is exaggerated in my opinion.
From a different perspective I've seen this many many moons ago when CNC machines became feasible in the cabinetry industry and everybody saw doomsday coming and cabinetmakers lose their profession. And yet here we are and those who implemented the tool thrived. Even those that didn't are still kicking and doing well.
well explained. thank you👍
Since AI is just a tool and not true AI then it isn't likely to replace artists, but it will probably become a popular tool used by non-artists to circumvent employing artists.
I agree that it won't cause Massive Industry change, but it in terms of anything other than the professional level, AI will take over.
It produces 'good enough' images and video that will drown the internet in scum. It's capacity to produce mediocre images at such quanitity is still worrying.
Just a plug about actors looking different throughout the movie... ever seen T.Gilliam's "The imaginarium if doctor Parnasssus"? ;). Great video btw! Lots of solid points.
Something AI misses is how specific you, yourself, can get with your own art. You can perfectly design everything you want in any way to the pixel level.
Even with stuff like controlnet, loras, everything, you can't change it down to the pixel.
All of the critiques like this seem to ignore how new this technology is. If you're saying "AI can't do this" you really need to tack "yet" onto the end of the sentence
Welp
if yer a kinda guy who says it will replace Artists- extrapolate your carelesness up untill it reaches your forhead and you get that if you let this happen once it will happen to you sooner or later
No empathy for someone's job means noone will empathise will your job once you are in danger
If AI gets advanced enough, I think our jobs might change from doing most of the process manually like we do now, to us become a type of lead 3D artist overlooking a team of AI. For example, we might use AI to generate a model, retopologize and generate a texture. Then we would step in to polish, fix and change what the AI outputs to make sure it's consistent with the project's style, and technical parameters.
Even though image generators can make artworks that look great, entrepreneurs and producers still hire artists to overlook the creative process since our expertise is not just knowing 3D, but also being able to take the right creative choices for the job.
I think that 3D artists not using AI in their workflow might fall behind 3D artists using it.
@@3Digitalist maybe. This video is only really addressing AI diffusion models as they stand. I don't know what will happen in the future.
Rob I've seen so many explanations of diffusion models but realized I didn't truly understand it until your explanation. You have a real talent for breaking down complex topics!
Thanks. I was hoping that explaining how it works would help to demystify it a bit.
All the things you say it can't yet do, how long do you think it will stay like that for?
@@RazorbackPT I don't predict the future
@@DECODEDVFX😮then your doom to not be able to live in the future😮 you never thought of it so it will be incomprehensible to you how is it possible I knew AI would be here around now I know exactly where we're going😮 or much more than you do because I'm paying attention you are not😮 you don't know about brain organoid computers😮 you don't know that right now a guy is being used to map all the neurons in his brain with AI😮 you don't know a new AI model is coming out that has 300 million contacts length😮 I do I can see and predict the future the future is happening right now😮 you're literally just talking about technology from a year ago and complaining about it😮 you're doomed your whole your people you're doomed😮 you should be telling people AI will take their jobs and they need to prepare for it you are literally making people starve in the street you in a way you're killing them😮 you're killing them with the LIE of Hope😮 multi-billion-dollar multi-trillion dollar corporations they don't give a fuk about your copyright your your artist integrity if AI makes more money you're doomed and AI does make more money😮 AI will continue to make more money and if they need to make a nuclear power plants and you and you have to live in the dark with no electricity to power the AI so be it that's what they'll do😮 you think because you find it morally unacceptable that it's not going to happen😮 what are you people children
You end every sentence with an emoji. I can't take you seriously.
There is no future in which AI replaces artists, unless that future is AI killing all humans. AI is not going around smacking the paint brushes out of people's hands. Even if all work is automated and we have nothing to do all day someone will be making art, in fact I would bet more people would be making art. It's one of those weird things humans just seem to do, like its baked into our nature.
Will it replace some (or many) forms of commercial art? Yeah probably, but I see that as a good thing. Work culture in the US is in severe need of reform and the arts are the best place to start having that conversation.
Well this is what many people have said: that art will return to being a rich people’s hobby.
Most people simply won’t have enough leisure time because they will likely have to work multiple jobs to pay their bills.
people will lose jobs tho...jobs that they are passionate about. That's not fair.
Similar issues exist in the world of software development. All AI generated code does is save me the effort of wading and sifting through scores of stackexchange posts. I still have be able to look at it produces and determine if it is correct. Even if it is correct, it might still not be usable directly. Will AI negatively impact jobs? Yep. There are people who will lose their jobs. Is everyone going to lose their jobs? No. Skills, thinking and competencies are still needed. AI is just another tool.
Yeah. I'm not much of a coder. I used chatgpt to help me with python stuff. The code is quite often nonsensical and when it works, it's often bloated and inefficient.
It provides a decent starting point for a novice like me, but it can't replace clean code written by professionals.
@@DECODEDVFX AI often is very good at giving you either a basic structure for something or sometimes also good at outlining bugs or debugging things, especially when you are mentally stuck on one angle. I used ChatGPT recently to generate a UI component in a framework I wasn't really familiar with which I just had to rework and read up on it - that likely saved me about 50% of the time of having to figure it out from scratch myself. That's an extreme example I would say, but it does boost efficiency and the increased throughput in workload means we either need more work or people will lose their jobs.
Coding is a really interesting case, since according to a survey, the majority of programmers enjoyed using ChatGPT in their workflow, however at the same time programmers using AI tended to have worse code. Like ChatGPT works great if I just need to write an simple addon for Blender, but I bet it falls apart when developing something more complex like flip fluids
"Just another tool" ... we will see if you change your tone in the coming years.
7:32 - There come in those miserable bastards who look at themselves as "AI artists", because they have mastered prompt giving to those generators.
AI Music is scary good, but it doesn't really give a crap what you put into it if you provide it a sample, which makes it feel like it's just crossfading someone's existing work, not easy to tell when that happens though.
Well I wouldn't call it good music, it's great at churning out tons of generic garbage the elevator music type stuff but I have yet to hear an AI song that had any real heart or that moved me in any way.
@@aegisgfx It sounds like something, but kind of like image AI sucks at making hands and feet, AI music also has some weird issues that get in the way of writing something that'll move you I will admit that. But music does seem like an area AI does a little better in - Unless it's just a hoax for freelance producers who take peoples requests and just slap something together in minutes, idk lol
The thing all seem to be missing about the AI debate (seen many on RUclips chime in on this) is the historical context. The conversation currently going on about AI has been made many times before, every time tech and innovation affects art. One clear example is the camera. Photography was even called by the famous poet Baudelaire "the refuge of every would-be painter, every painter too ill-endowed or too lazy to complete his studies". Basically the same quote is said now about AI. Just a few years ago, the same was said about people using photobashing for concept art. This is just the same debate going on over and over, and I barely see anyone on RUclips mentioning it. This could also be said about the whole "AI is not art"-statement some people claim. The discussion about what is "art" has also been going on for a long time, but still no one brings up important works like Duchamp's Fountain in these types of videos.
P.S. Not necessarily claiming the video argued all of the points brought up above, just saying history is a nuance which is missing in the general AI debate.
From what I've seen, this historical context is exactly what AI proponents often bring up at one point or another.
I would like point out that the camera, while absolutely as revolutionary and upending to the art world at the time as AI is now, was developed in isolation of artworks (although it more or less built off of technology that artists had used for realistic renderings previously). The same can be said about digital art software, which offers an alternative -- and often more efficient -- approach to traditional processes. Generative AI for images and other forms of media, on the other hand, could not function without the millions of artists and billions of artworks that (unknowingly) stand behind it. The generation of work from a simple text prompt is completely devoid of process, and is more of an exercise in vocabulary and syntax that is ultimately guaranteed to return good results than it is a tool.
Tldr, generative AI is not a means to an end like photography or digital art programs. It IS the end. It does all of the work for you, and it could not function in absence of reference material.
@@OctaveIndustriessmall correction - generative art doesn't need art to be trained. They have been now but it's not necessary.
@@OctaveIndustries Your comment unfortunately makes little sense to me. Photography is the press of a button in its simplest form. How is that process more advanced than writing a prompt? The core difference is that photography now is accepted as an art form (but it was not during the time of Baudelaire).
Photography, in its simpler form, is a lot about having the eye - finding interesting shots and capture something. Same about AI. To make interesting art with AI you need the eye, to direct the AI into making something valuable.
And if directing is not an art form, is a film director not an artist? The film director does not do the photography, or the acting or the writing (usually that is). He/she directs. And what about a VFX supervisor?
The issue with the anti-AI movement is that it doesn't make sense if you judge AI art the same as previous art.
@@OctaveIndustries exactly. people can claim that coming up with prompts is 'just as hard and requires skills', but at the end of the day, the image they generate isn't created by them, but by a machine. Imagine giving an idea to a painter to draw, and after seeing the end result claiming that they were the ones who painted it. It's the exact same thing.
This ^^^ ‼️💯💯
I believe AI will add more value to real artists that de value them in the long term
Totally agree that people who want to do VFX should do it. Disagree that AI "cannot do things" based on current state.
Also, I don't think you have mentioned that you can use AI to "improve" sketches (either be line art, or more developed color drawings).
One more thing. What's the difference between AI learning from other artists or a person doing exactly the same thing?
Okay, start learning from the works of Rembrandt, or Vereshchagin, if you don't want to live and want CGI, start copying wlop or any other artist. I'm sure your result will be different. And AI will identically steal pieces of work, but in fact this is not intelligence at all, it is marketing that made it intelligence. + Using the works of artists when learning AI is illegal because you are literally using someone else's work to make a profit. I am an architect, I am happy with the topic of AI development because it simplifies routine tasks, but it cannot sign papers and bear responsibility) But I am very afraid of the issue of devaluation of art, because it can lead to stagnation. Art is what from early times distinguished a thinking person from a monkey. I hope the value of human art will not fall in high circles, but will only increase, and that the development of AI will not lead to the fact that most people will become like monkeys.
@@denkot442 Why do you say it is not intelligence? How are you different from an AI?
I answered your question, look for the difference between machine learning and artificial intelligence yourself
@@denkot442 I happen to know quite a bit about that.
Well then I'm very happy for you that you came to the conclusion that you are no different from the AI after your research.
Sadly, the tattoo industry is flooded by this.
Yeah, I can believe that.
I already quit art last year.
@@astrea555 :(
...why?
It's seems there's a significant knowledge gap in what you think AI does and what it now can actually do.
With multimodal AI almost all of these issues are resolved/going to be resolved in the next 6-12months.
Besides, I've been thinking about the differences between AI and humans, has there ever been good artist who never in their life copied someone else in order to learn?
@@Think666_ Copying to learn is fine. I've published master studies many times on this channel.
Copying and pretending it's new is a different thing.
@@DECODEDVFX It reminds of that saying "Great artists copy, bad artists steal".
I don't think it's about copying, but being able to take inspiration and using it in a new way. Since it's almost impossible to make something completely new.
However if I used AI to generate a picture then post it on Instagram and saying it's made by me, that would be wrong. However then again, the client does not care about that, they just want good results
I don't want to be that guy buuuuut... a lot of the flaws and weakenesses you mentioned in here are being worked on right now.
I'm not an AI artist but i've been dabbling with it for some time and people are finding very creative ways to work around those flaws and getting better at using AI
I think that you're right talking about AI now; but this is like comparing Blender today with the first 3D program. AI for art has just started (less than 2 year ago, I believe). We are just at the very beginning.
Also, I don't think you can say that AI doesn't understand "form". Clearly it does, and it is getting better. To me it is like looking a drawing by a 5 years old and saying that humans don't understand form.
Yes, AI will replace specialised artisans completely. But it will offer everyone the tools to make their own creations.
If you want to watch a movie with Charlie Chaplain, Brad Pitt and Elon Musk in it... in a musical remake of Casablanca with a happy ending.
AI will make it instantaneously for you to enjoy,. Or in 3D, in a virtual headset... with you as the main protagonist.
But the most important aspect, is that AI will teach itself. It will create new forms of entertainment and 'art'
It will free people from the laborious task of making stuff... to being the producers of your own entertainment.
Remember, with AI...your watch ( if you can afford it ) will be your doctor. A career as a doctor...will be over.
The AI enabled future will revolutionise everything.
They said the same thing about Cameras replacing landscape artists
Artists still exist
And if AI replaces everything
Wouldn't everyone become dumb?
@@SMCwasTaken Landscape artists are hobbyists. Just like knitting... it's a pastime, a repetitive task in order to disappear from the world and yourself.
I just can't get enough of the fact that nobody cares about stock authors and for some reason using stock database is justified. Why so? becouse authors agreed to these? It;s true, but nobody asked them beforehand! They just got it as ultimatum. More so, once you agreed, you can't opt out. They will have all your images, photos and illustrations embedded in AI model, forever! Oh, and you also will not get any of payment, so.
Just leaving this comment so I can come back in a few years when AI has replaced the jobs everyone says it could never.
It can replace doodles, but it can't replace animators, 3D sculptors, texture artists, etc.
The biggest danger is that it is cheap and both publishers and the audience don't give a d*** that it's crap. It will become a "style".
Yup. Look on game developers, most of the giants had been releasing rehashed and reused poop for years now. If they gonna get things done quicker and save money they gonna do it and use AI.
@@dexterdexter7923u will play a game with a tonne of glitches and malfunctions.
you basically described, how humans learn. Children go to school, teachers show them renaissance pictures and tell them this is a renaissance picture until they can recognize them quite well. That's also how you learn new skills. first you might make many mistakes, but you train again and again and later you will use your skills very well. And retraining is difficult for humans too, once you learned something and used it for years, it's hard to use something new.
Just saying...
And as concerns to copyright: people may know the Mona Lisa and then they start to draw these images themselves. Are they violating copyright laws? According to your logic yes. Basically as soon as you see any image / painting, you need to buy a license. right?
You have a handle on it? You know what it can do? Those are some bold statements. Do you know what it’s going to be able to do next week? Because That’s what people are worried about. Not this week but next week and the week after. And next year!
The problems with the current AI models are inherent to how diffusion models work. They can be improved over time, perhaps, but they can't be fixed.
Future models may solve these issues, but right now, this is the best they can do.
Nope. Ai doesn't have life experience and feelings, just knowledge. He's like an academician, not a creator/artist. And human pretending being artist using Ai has just prompts to proove his productions. Real artists have original éléments, sketches, roughs, researches etc.
99% of those artists have terrible ideas...
@@comebackguy8892 probably ;) but my tone was drastic. I am not criticizing those who only use AI. These people may have fantastic ideas but not the talent to bring them to life. And conversely, a talented artist is not immune to running out of inspiration.
Definitely!
AI will make it easier for us to do the projects we wanted. Even though anyone can generate an AI image, artists will still be needed to make the results actually good.
It's like everyone can take photos with their phone, but we still admire great photographers
Everyone talks about how AI is stealing work. But artists also “steal” other people’s work for inspiration. An AI can have so many pictures that it takes from all of them and isn’t necessarily copying one. There are exceptions, like that girl he mentioned, but I think it would be fixed with more data.