Will AI Never Become Legal? Photoshop AI (2024)

[@BoroCG]

So it seems like people on both sides of the argument may have missed the point I made in the video about the legality of AI. I think most agree that AI needs to be trained ethically to be useful. But the problem is that it seems to be impossible, due to the nature of how AI is trained. This is something I recently realized, and I think it's important to have a conversation about it. It's just too many images used to train AI to be confident everything's clean. Even if all images were used after the approval from authors, and even if they were paid for it - who can guarantee that every single one of these gazillions of images are legit clean? What if the author stole an image? What if the author used AI to create their image, and then never told anyone? What if the author didn't even know they used AI? When you're using generative tools, you're blindly relying on the huge amount of copyright cases, just hoping everything's actually clean, and nothing will come up in the future that will make all the pieces created with this model pretty much illegal. As I said, I like AI as a tool that can elevate the scale of an artist's work. It has a potential to let one artist make a AAA game or a feature film. I said a few years ago, many artists would become directors this way. But all of it seems to be actually impossible, considering the legal vulnerability of any AI model, even the most "trust me bro 100% legal" one. The only way to make it work that I can think of is just hoping the tech of AI can be improved enough to give good quality from a very small amount of training data. I would love to be able to train AI on just my own artwork, and just use that as an army of art-clones of myself to quickly build something huge. But for now, AI is in it's "illegal" stage of development.

[@BoroCG]

And to those who criticize even the fact I'm showcasing AI in my video: I don't get that at all. I'm showing it to let people know about the state the industry tools are in, while having a conversation about the legal problem of it. AI is not a sin, it's a technology that needs to be developed to be good. We all know AI is not going anywhere. If this conversation doesn't happen, then AI can only develop as evil.

[@invertexyz]

You list a bunch of "what ifs" about ethical data, and it just ends up sounding like because you don't see an ideal solution, you're just giving up. Nobody is really claiming we can completely avoid unethical data getting in. That's not the point, the point is to do our best to stop as much of it as we can. Your argument is like saying "well, we can't stop all crimes, so let's just get rid of the laws and police.". We already have legislation being pushed through to require companies to show their training data. That would go a long way to solving these problems, because people would be able to reverse search their content in the dataset to see if someone else submitted their work. This is the kind of legislation we need advocated for around the world. And if these tools end up only receiving mostly regurgitated data from other AI tools, then the results are going to be majorly affected. My criticism is that it's not a great time to be mostly promoting the use of the tools while we're in such a critical stage of shaping public perception and advocating for laws to help benefit humans and culture instead of companies. At the end of the day, public perception is the most powerful force society has. If most people end up agreeing that using these genAI tools is bad, then most people won't be so inclined to use them.

[@BoroCG]

@invertexyz, what are you talking about? Why are you assuming I'm proposing to give up on laws?? I'm describing the problem with AI, and that it can't be used by any artist for actual work in its current state, where so many images are used for training. Please don't assume that I'm your enemy because I don't necessarily hate the technology of AI. ' Your argument is like saying "well, we can't stop all crimes, so let's just get rid of the laws and police." ' - No. My point is that we can't use AI.

[@marsmellon]

2:56 Adobe actually uses pictures from users who use their programs, so they're not really clean. They take stuff from Adobe Cloud. It's a hole thing: Many people lost their trust in Adobe and stopped using their programs. I get that this is an old video so maybe it wasn't know when you made this.

[@gtbgabe1478]

Someone on Tumblr explained it the best: "I always wanted AI to develop so we could have robots doing our boring tasks, like washing the dishes, so we have more time to do fun creative stuff. But they are developing it to do fun creative tasks and leaving us to do the boring tasks like washing the dishes."

[@ianalexanderreyes5890]

actually you already have a dish washer, tho its very expensive

[@floriakson]

I'll try to say things that have mostly not been said already in the other comments, no "grrr AI evils", just facts and I'll try to actually answer your points, because I really like your work and you seem to be open to discussion in this video, so I'll do my best to explain *why* there is so many "grrr AI evil" The fact is, your whole video demonstrates the case about not using AI - The legal problem is not some "gray area that will always be muddy", it's very real and inherent to all this wave of generative AI, you can give it a thousand pictures of a person's back and a thousand picture of a person's front, it will never figure out a profile because it does not create. It's a picture recognition algorythm that analyses patterns, synthetises pattern, and reconstitutes patterns. It's being *marketed* as a creative tool but is not and will never be, it's litteraly a machine to replicate mixed and mashed pictures. - Yes, It cannot exist without an absurd amount of training data. That's precisely the defense open AI had against the claims of illegal use of stolen image : "We couldn't have created it without stealing so we did". That's also why every attempt of *seemingly* ethical training is miles behind the competition, the gargantuous amount of images necessary to train it is not required to improve generative AI, it's fundamental to it being of any use at all. It can only create images that exists. That's also why... - ...the above problems will not go away. It is not a matter of the technology being young, it is the very *basis* of technology. Very real, very recent concerns about the development of AI is that it cannot improve by using pictures generated by an AI. It makes the result degenerate and at best it kills any variation in said results, making it an absolutely useless repetition of the same results over and over. (After all since AI creates synthesized patterns of prompts, the synthesization of a synthesization of a synthesization will always be the optimal way to anwser a prompt). If nothing is done to solve this blocking point, of if there is no solution to it, the whole model of machine learning generative AI *will* need a constant amount of man made/taken pictures not just to improve, but to MAINTAIN itself. And with AI generated images being sneakier and sneakier, it will inevitably slip in other AI's training data. If everything stays on it's current tracks, AI will litteraly kill itself in the next years. (all of this without mentionning other messed up stuff like the fact that AI is not AT ALL a magic tech stuff but using the very human labor of thousand of exploited people sorting hundred and hundred and hundred images to make the services viable. There is no magic "please don't create gore" button. Actual, real people had to watch thousand of gory results to say to the AI "don't do that". - Throughout the video you mention the adobe subscription paywall, and how it's weird that you're training AI so you can pay to use it. Yes. I'll let you rethink that knowing what I've mentionned above about the constant influx of man made art necessary to maintain it. But I'll also warn you : It will get worse, adobe subscription is not even the beginning of it. The current AI services are not profitable. They're accessible because they need people to use it to train it (remember that every AI service open AI released was a training version to collect data, not an actual product), but at some point they'll need to make money with it. They already do, that's why there's an adobe subscription, and that's why the open access chatGPT is still generations behind paid chatGPT. Especially considering the huge investment made currently to bruteforce the new features and keep the "new" feeling. It's an economic bubble that stands on a buzz word, it will burst, it'll get unusable or absurdly expensive, or both, and artist will have suffered because of a bubble. - now getting into the practicality of using it : you video demonstrates... that it's not practical. You said it yourself : there is no control over it, or at least certainly not the kind of control needed to make anything artistically useful. No way to make actual, informed art decision beyond replicating pre trained styles, because it's not made as a tool to help the process but to replace the process. You're supposed to use the prompt, get what you asked, go "wow" and voilà. Could machine learning or algorythm be useful for art ? IT ALREADY WAS ! that's what blender renderer's denoising was, photoshop already had the patch tool for decades now, the current gen of generative AI is doing a way worse job, with all the problems mentionned above, without any control, and in the least effective way possible (both as a process and in the way it's developped with so much ressources and money to HOPEFULLY anwser correctly to prompts. Wich it has trouble doing.). To be perfectly honested I'm kinda bothered by all the things you "let slide" with how it responds to your input just because it's AI : it doesn't follow your selection, it doesn't get what you want it to do, and it does a very poor job at extending. If someone presented the same picture in a world where AI and it's marketing doesn't exist, it would be considered absolutely terrible job. Why ignore all the problems it comes with because it could be useful if you also have to ignore it's not useful and probably cost you more time for worse results than just browsing google pictures or photobashing ? The very people that create AI are targeting it to people who do not want to create art but want art created for them, and do not care how or at what cost, it could be used to create tools, it's used to create gadgets and gimmicks. Actually I think I get why you seem biased and enthousiast, not just in this video but with your grandma's design for example, but I don't think it's a good bias : it's the same bias as tracing art, your falsely happy with what you produced because you "made" it but it does not come from you, so you appreciate it as if it was another's picture while also thinking "hey, I did that" (TO BE CLEAR : I'm not accusing you of anything, I'm trying to interpret why there is such a disconnect between how you appreciate the things you do versus the things you use AI for, not just you but AI users in general) - who is AI useful for ? Everybody is tired of seeing it, because our eyes are pretty clever organs and either we see it and we're tired of the AI look, or we don't and the whole point is to hide the fact it's AI to decieve people. Numeric painting wasn't made to pretend someone drew, it was to get more ambitious. CGI wasn't made to make you believe someone made a stop motion animation of a t-rex, it was to create imagery never seen before. Is it to get new, fresh ideas ? AI litteraly cannot do that ! in your video it creates a beach. The most basic, cliche beach you can get. The most optimised anwser to the word "beach". How many more unique, more interesting beaches exist ? How do you know if you trust the prompt instead of actually discovering it through research ? How many actual fresh ideas are you missing by letting AI confort the biased idea of what each prompt should look like ? It's more work, it's troublesome, but let me ask : We've already got plenty of useless pictures flooding us everywhere, what's the point of more pictures that no one cared enough to actually create ? Maybe the real question is not "Why shouldn't I use AI to create this picture I'm not willing to put hours into making", but "Do I really need this picture ?" So... what's an usecase I would actually be happy wih considering AI ? Currently none. It's gone WAAAAAAAAAAY past it's actual value being a funny tool to create pretty pictures, and the more is goes on the more stale it stays, while accumulating more and more and more and more and more problems. Where do I think it WILL be actually used when all the novelty will wear off ? stock footage. Illustrative media no one really cares about making, no one really cares about looking at it on it's own, but it's still necessary, it can elevate a production, it can be reused and reinterpeted (that's where you get great memes). Hey, at least with ai stock footage it could be a bit more customized for each production. TL;DR : "But if you consider all the problems it's only usable for memes - *Chad yes*" Have a great day nonetheless, and I'm still very much looking forward for progress on your game !

[@nikicherry1234]

yes

[@ianalexanderreyes5890]

do you know what inpainting is? you can generate, let's say, an arm on your artwork that doesnt look good and fix it with AI, after a few tries you may get decent results, specially with the latest stable diffusion models, so yes, it's a tool and very useful

[@EmilyE96]

Your experience is kinda what I've had with all AI models - it's actually really frustrating. As an artist, it's so much easier to just paint/draw (even photobash), not rely on the roll of the dice. As you said, you're giving up a lot of agency.

[@thejagman22]

There is one thing that might interest yourself or your viewers if you're not already aware. You can make a rough drawing to feed your Generative Fill results. For example, in the case of your egg, start by drawing out the shape and colors using regular Photoshop brushes. Then when it comes to making your selection switch into Quick Mask Mode. For those that are unaware Quick Mask Mode allows you to literally 'paint' a selection using brushes or other tools. Black swatch paints a selected area, white swatch removes from that selection, and adjusting opacity creates a semi-selection of that opacity. (Sidenote: I'd strongly advise double-clicking the Quick Mask icon which opens a dialogue box - switch the defaults from Masked Areas to Selected Areas, additionally I'd advise changing the opacity in this dialogue box from 50% (default) to 100%). Then use a normal hard brush (or soft brush, just be aware that the soft edges might affect outcome of results), set to say 50% opacity (the lower the opacity the more consideration Generative Fill will give your guide drawing) and paint over /cover the entirety of your guide drawing using black as your color swatch, which will display as red in Quick Mask Mode. You can then switch back into normal mode. Note you may not be able to see your selection as "marching ants" due to it being a semi transparent selection. Next step use the Generative Fill, with prompt as normal. What this process does is takes into consideration the shape and color of your guide drawing underneath. You'll find that what you generate resembles more strongly your guide drawing and therefore removes a substantial amount of the randomness from this process. You can of course experiment with different opacities, brushes, etc to fine tune results. This same method also works if you begin with a photograph or any sort of image and takes more cues from what's already in your image. Hope that helps in some way. It's something I use personally all the time.

[@efs3]

Honestly I think ai had potential to be quite useful but the way it's being used is just so useless and unethical. Originally I thought Gen ai would be used to generate photorealistic concepts like in Photoshop but then people started using it to just make "art" that looks like someone else's, just to feel good about themselves. Art mimicry is like a huge issue with it and I think the direction it is mostly going is really bad.

[@ArtisanCris]

Not only that, but people are making porn of real people with AI. And not just adults, but also children. It's unethical AND criminal.

[@BadTunes.]

"It's like for memes " IS THE BEST way to discribe ai art rn, Btw i think ai will improve more and more no doubt about that ❤

[@Henqueri494]

14:15 - Silence guys, hes discovering capitalism!

[@BoroCG]

I just find it silly how every big company enters the AI market with literally the same product of "you type in text, you get an image out!" - and starting with the same poor quality, when we already have much better stuff elsewhere. With other tech, there are patents and technologies that can be used by all companies to move forward together, not just redo everything from scratch every time

[@vogelstrauss9]

The problem with AI is that it doesn't solve anything, it's already an obsolete technology in the artist field. I chose to be an artist because I enjoy the process and everything that goes into creating art. AI image generation simply takes that away and replaces it with instant gratification. To answer your question: I really don't care about AI and even if it's completely ethical and free of any corporate greed, I'd still not use it, because it's just so uninteresting. Pressing a button and wow you have an image...what now? Repeat the process? Do it for the next 50 years of your life until you can finally retire? That's just not it for me. It's a cool toy for your average consumer but that's about it, really hate the fact that artists have to put up with it just for that.

[@simonspethmann8086]

I've seen artists use it for reference creation, which makes the Photoshop integration quite handy. You can Frankenstein and refine (well "photoshop" 🤣) the AI images, photos, your drawings and end up with something useful. As a reference. It's essentially like photobashing, just easier and faster.

[@dorum358]

I love how calm you are when talking about issues like this, it really makes it easy to focus on your point and get over all of the anger and passion associated with the subject. I'd like to make a few points that might interest you, it may be a bit long, but I think it's interesting. Also, everything I'm saying is related to the industry where I'm from, it probably differs in other parts of the world. As a professional artist and animator working in film and also advertising I've noticed AI being integrated into some productions, or at least them trying to integrate it, mostly not as the last unedited version of that certain advertisement. It's usually briefs or storyboard-like documents that were normally done by those employees from advertising firms whose only jobs were to make moodboards from pinterest and stock images and pitch them to their higher-ups. Those guys have moved onto using AI to make those briefs and us artist that receive them have to try and translate them to understand what they want. I have also received images that were generated with AI and when they tried to make it edit them it failed, so I was hired to manually paint over those. In some situations they failed to get a proper characted design for the AI, so they hired me cause it's still easier to communicate with an actual human artist to get the changes and feel they want haha. Also with the style I have, it's always faster to just draw stuff by hand and get exactly what I want. I feel like that would be a similar experience for you too. I also think if it was possible to train a model with only my art it would just keep it exactly at the progress level I was when I trained it. So it just feels like a one time thing, any evolution I have in my style would require making lots of drawings and retraining the AI, which sounds like a hassle, it still feels like there are easier alternatives to get to the result I would want. Only thing it feels very useful for is realism, which was already done in the industry using photobashing and lots of stock images, so there's really not a huge difference there either. So far, from what I've seen in the industry, I don't think any reputable brand wants to risk using AI as the final product, cause they don't own it, which is exactly what you described, but unfortunately sometimes some things slip through the cracks. As it is right now, conceptually, it's (hopefully) never going to be ethical and legal, but some people unfortunately just don't care about that. A half-alternative would be a system in which if an image from the training data would be discovered to be stolen, it could be removed, the AI automatically retrained, and the original owner compensated, but the content that was generated based on the training data is also compromised, therefore this solution is still a dead end, not to mention the technological cost of such a thing. As it is right now, to at least sort of "prepare the ground" for any development for an ethical system, we need a different copyright system for training data. Even if a company owns an image created by an artist, they shouldn't be able to train an AI with that image, that should be a separate right that the artist has to agree to do and be paid extra for. It's like buying artwork from someone and then training a model to be able to perfectly replicate their current style, if you'll render those old works obsolete at least let them get paid way more. It has to be opt-in, and also be different from the current copyright system, because it's a new technology that completely screws over the old system. Also, the biggest point I have right now based on my experience is that this technology currently benefits big companies more than it helps small artists. It's an excuse for greedy corporations to fire more people and the opportunities it creates also mostly only benefit the big guys. Like, think about it, the big brands that have money only want to work with big advertising firms, cause they are a safe option. Those firms then fire artists cause they can replace them with less people that use AI and those other artists are therefore out of jobs and no one will hire them because the big firms control the market. Also, in film and gaming, if someone wants to make something big and has an actually good idea, they don't need AI to make that, they just need to pitch it to someone with money or do a public fundraiser, that system already exists and it benefits multiple people, not just the person with the idea. If anything, we shouldn't try to make AI that replaces the need for people, we should make it easier for people with good ideas to be able to get funding and form an actual team.

[@invertexyz]

When the data used for the AI is properly knowingly opted-in instead of broadly scraped (or ToS change forced opt-in without asking us), then maybe the discussion about okay workflows could be had. But that's not the reality yet, so I don't think it's great to be even giving these tools positive attention while they exploit the world's artist. We're trying to fight to get back some artist's rights, not contribute to people's acceptance of it. Been weird seeing you make videos using it to be honest.

[@ivanmatveyev13]

Are you new? He made a video arguing art is dead and he is shilling ai for years now. How can you be surprised?

[@nestorlichev8531]

​@@ivanmatveyev13 damn you are the master of gaslighting. What a dumb, spineless comment.

[@idkwayta1722]

Isn’t the adobe stock tos already warn people about this for years? And all adobe training data is licensed?

[@invertexyz]

@@idkwayta1722 The licensing terms didn't state things about AI until recently, and is not something people would have had in mind when signing existing agreements with websites. The way these AI tools utilize the data warrants a re-issuing of accepting new terms. These big companies also use their power of knowing people don't have many alternatives to essentially force them to sign these agreements otherwise they can't use what they need to to live, but then their data is now being used to try and get rid of their job... All the "licensed data" companies are claiming they're buying is literally just scraped data that is repackaged and sold as licensed content, shifting the liability to some other company. Adobe's already been caught on that as well.

[@BoroCG]

@@ivanmatveyev13 hey, I never said art was dead. I said digital art, as in digital painting, has never really received the impact many artists including myself expected it to get many years ago. But that's a completely separate, almost philosophical question that had nothing to do with AI.

[@ThecatThecat-hq1op]

What you're asking at 14:04 is pretty much "why does competition exist". If there is only one product that everybody *has* to use, then there is no incentive to improve that product at all. Usually what you see is that when a product becomes a monopoly, it will get even worse as the people behind it start to cut unnecessary costs. Also having different products created off of completely different design philosophies can give a lot more meaning to your choice of what to use, and it will guarantee that one of those design philosophies are future proof. If you're looking for a more collaborative environments with better opportunities to be compatible with other stuff, then you should look towards open source projects. Though the problem with those is that the people behind them don't usually make money, so progress can be slower. An open environment, with motivation to get better, a lot of options, and a way to achieve personal security/growth for the developers, would probably produce the best products.

[@BoroCG]

Yeah you're right about the competition and the fact that things are built differently to try and find better ways and all. I just feel like most tech giants are motivated by very different reasons. It's a race towards getting a god-like status with GAI, and it seems like it's all that matters to many of them

[@jonaseriksson3782]

AI art today I think is just used as a gimmick mostly. However, given time and interest, it could be used as a tool for creativity instead of a point and click adventure. There are plenty of menial work associated with creating art that just serves as being cathartic but not necessarily artistic. For example when editing a photo we have tools for removing a background, which is already an improved tool over removing the background by hand, and that could be further improved with AI. I'm sure there are plenty of creative souls out there who can't draw and have no real interest with it, but if AI can draw something for them then maybe they can utilize that further within the creative field they enjoy (like storywriting, or animation). I think it's understandable but close-minded to completely write off AI as uninteresting or "cheap art". If we don't learn to adapt then we will eventually become dinosaurs surpassed by everyone else. Personally I find myself being often cynical and annoyed at AI artwork, but that only comes from me being scared that people who do not draw won't understand the time and dedication it takes to learn it and present a single artwork whereas that intricate AI landscape painting probably took two sentences. AI art isn't going away and pretending it doesn't exist will only make you a grumpy person.

[@dietrichwolf8189]

i agree with you for the most part on your point of view regarding ai (like also what you statted in older videos). would be nice if ai would be at least freeware if the artists don't get paid for their participation and a full disclosure on what was used and how for the public.

[@origamiPL]

the recoil from ai on all of the companies in about a year will be the funniest shit ever

[@verdedoodleduck]

Thanks for the walkthrough! I thought the software was quite disappointing - but I think, while they were obviously using an old model and it is inferior to what is currently out there, it parallels every experience I have had with the ai bots. I think until we get to an general artificial intelligence, all the images will be 'good enough' for a lot of purposes but 'not good enough' for most specific purpose. I doubt it will be common even in the near future to get exactly what you want but you'd get something that will be fine (most people won't even care and of course many times the images are cooler than you would expect.) I'm guessing that somehow at least the big companies will mostly get away with what they've done and are going to be doing (with the creative works of others that they are bleeding for profit.) so probably some of those images are safe - but that's just a guess...things could turn sideways due to copyrights.... The current generation of ai models will always lack intent or understanding of the meaning of what is being generated. It's fine if you're happy with the product but if you have something specific in mind you are still going to need a human that can draw or write, etc. Still, our days are probably numbered and even fewer lesser artists (such as myself) will be able to make a modest income from it in the meantime. :\ After we get the AGIs going then maybe art becomes something really niche like ordering a hand-woven sweater. Of course after the machine uprising an art career might have more promise (probably be a high demand for clever survival signage).

[@nkmvkclhvfe]

Great tutorial!!!

[@Aerderranissar]

You should try new AI addon for Krita. As you know, Krita is free and so is addon which uses Stable diffusion. It performs much better than photoshop AI if you have somewhat decent graphics card and you have much more creative control over the process (by choosing your model, lora etc.). Can even train your own model on just your art. And it has real-time preview as you describe at 6:45

[@ThecatThecat-hq1op]

This is the closest to my ideal implementation of AI that I've seen. I just wish it was better, more real time, had more tools to control it, and had different things to do with it beyond just generation. I would also like it to be FOSS, and run locally on my own machine without any cloud involvement. No subscriptions, No adobe.

[@wolfboi8785]

who agrees to train an aspiring artist? If you got into art because you loved comics and wanted to make your own comic books, and therefore looked to your favorite comics for inspiration and references, do you suddenly need to pay the original artist? do you need to get their permission for this? And im not talking about plagiarism which is a term which people throw around willy nilly when talking about AI, but im talking about creating an original creation but using those previous artists works as inspiration or refrence, as this is exactly how AI Works. It does not "copy paste" it doesnt steal someones art intentionally, it takes inputs and is taught what those imputs are and then produces an output. This is exactly how art school works, you go to a school, look at art like what you want to make, learn from it, and then use what you have learned to create something new. And sure some people might use that knowledge nefariously and copy and plagiarize other artists, but the majority of people are not doing that, they just want to make something unique by using information gained from studying other peoples work or nature. AI is exactly the same, we put in study material in the form of in the case of this analogy artworks, tell the AI through complicated coding what these artworks are and what elements make them up, and then through this the AI learns what those things are, then when you ask it to make them, it goes back through its memory of lessons learned and just like an artist goes "So that would probably consist of these elements, these colors ect" and goes through stages of creating the image. Conceptually there is no difference between the AI making art, and a traditional human artist making art, and practically there is no difference, because at the end of the day the "AI" is not sentient or even that intelligent, its probably less intelligent than an ant, Its just a tool that is used by another human to create art. And this is why you see so many AI artists using the analogy that AI art is like photography, people are upset and scared because a new tool has come around which threatens their livelyhood, but just like photography in history, I dont see AI replacing actual artists anytime soon.

[@rainbownotation]

It was weird seeing how it wouldn't even fill out the lasso tool selection you had made, which highlighted your point near the end about how much control you are giving up to the tech when you ask it to generate something. AI has a long way to go, and this sort of reminds me of when everyone was inventing the camera and film all at the same time, all starting from scratch but in slightly different ways and all ending up with something that is not as improved as it could have been if they had been working together instead. I would be interested to see AI tools made by artists, for artists in the future... but who knows how long that's going to take. Also unrelated but total mind bogglers for AI would be AI generated cooking recipes, and AI generated knitting and crochet patterns. It feels so strange to look at something that initially seems alright, but then on closer inspection you realize that it's all nonsense. Something masquerading as human and then turning out to be not that. This is why I find your uses of AI and perspectives on it to be interesting, because it seems like it'd be a rich opportunity to explore some strange and horrifying concepts.

[@artkonina]

12:40 ya I'm pretty sure a generation has a resolution of 1000px x1000 px... so that's why the egg was pixelated and was able to upscale but the dog was already smaller than that res so you couldn't upscale it

[@mezozoltan1]

The generative fill tool only generates a 1024x1024 image (at least it used to they may have upped it a bit), and the stretches it across your selection. Thats why it looks so horrible. One way to bypass it is to only select squares that are that size, which is stupid but it is a way. There are also plugins that do this.

[@DigitalArtcast]

Until the data set is ethical I don't think you should even be using it. Unfortunately firefly is still using the unethical set so it's not only illegal in some countries it's just bad practice as a high level artist to be messing with it when it's literally right now taking work from people

[@ff7522]

TBH the legality of photoshop's training datasets are near the bottom of my concerns for AI. And to my non artist friends, it's also the least compelling argument. "You used adobe ai on a small part of your texture and adobe used midjourney images which had a training set that might have gotten images without the artists permission". Ok, prove it? Who is coming to collect from you?

[@polycrystallinecandy]

There is a fundamental misconception that AI art is like photo-bashing, that it's mixing bits and pieces of original data. In reality, it's much closer to how humans learn by looking at other artists and the world around them. It can create completely new pixels that it's never seen before and it has basic understanding of spatial relationships between objects, the general concepts of faces, symmetry, etc.

[@ArtisanCris]

Press X to doubt. AI doesn't understand composition. It's a simple copy and paste machine-- that's why it makes weird hands every now and then. It doesn't understand that hands only have five fingers. It doesn't understand that hair isn't connected to each other, but is a clump of strands more similar to ribbons. It even gets eyes wrong more often than not.

[@polycrystallinecandy]

@@ArtisanCris They are probabilistic models with many shortcomings still. They will get things wrong. But to say it doesn't understand anything is being intellectually dishonest, even if it were actually a "copy and paste machine". If a human were given a thousand photos and was able to create paintings, even mediocre ones, based on any prompt by photo bashing, most people would say they had some level of creativity. Photo bashing is a well-recognized art form after all. The fact why AI is not photo bashing is harder to explain, but it's pretty simple to test. If you trained a model on humans and wolves, it already knows what a werewolf roughly looks like, even if it's never seen one. It doesn't know the word "werewolf" of course, but the representation is already present in the latent space, and you can find it by exploring the space between the vectors for human and wolf. This is a simple example, but this kind of inference/generalization is the whole point of AI. If you're interested in art at a philosophical level, it's absolutely fascinating. I've tuned models on my own work with pretty interesting outcomes, you should give it a try

[@grimsonforce7504]

​@@ArtisanCris LMAOOOOO neither do the majority of these "artists" many of the lack the basic fundamentals yet cry about AI. I honestly believe it's because they hate that a machine can produce far better work than they could. I won't even get into references artists copy paste as their own for profit. Especially with the rise of 3D models. Don't even need to think just copy and trace. I implore anyone to volunteer by reviewing art portfolios for even a week. Hell maybe even a day I guarantee you'll notice the severe lack in quality work. Yet the artists constantly praise themselves for smelling their own farts. AI is a tool nothing more if a machine pushes these artists to quit then maybe it's for the best.

[@ninthreaver]

I think like most technology it's a tool but using it, saying its your own work for profit should be unlawful. I would use it, in my case for inspiration.

[@ianalexanderreyes5890]

what if i render over an AI image? would it still be unlawful? how much AI generated image should be visible to you to be unlawful? how much do i have to change it to become copyrighted by its own?

[@mungeons]

I have a lot of major gripes with AI, exploiting artists is a big one. The AI would be nothing without the photographers and artists who’s images it takes from and does not credit or compensate. Secondly, people will just use AI to type in a final product which i find to be kind antithetical to point of art. However, I think i could care less if people use images from AI to do photo bashing or be apart of a smaller piece of art. Generative images should be used like all tools in an artist kit, like a tool that helps you get to the final product, not the final product itself. A lot like when people half hazardly used interpolation without proper use or intention just took a bunch of anime clips and used them to make a bunch of really shitty videos. (there’s also the whole sustainability issue with how much of a carbon footprint it makes along with how much money it costs to actually continue to run AI so the only way it can even be funded is by huge corporations throwing billions at it and government grants but no one cares about that lol)

[@Nurolight]

The way I see it, Generative AI kimages should be seen like a visual Library of Babel. Every possible combinations of pixels exists out there and when you hit generate, you're finding a singular instance. I hate the people that use phrases like "I made / I created" ect. because I see it more like they were the first to discover that specific image. Anyone with the same seed could find the same image. It's not truly a creation of the persons until they modify it in some way.

[@ianalexanderreyes5890]

how much modification is needed? 10 pixels? 50 pixels?

[@ianalexanderreyes5890]

also you will need the whole workflow to create the same image, not only the seed, that includes all the inpaintings you made, all the controlnet and extensions you use, all the upscalers if used... AI can be as complex as you wish, same with pencil and paper

[@maddyaurora]

it's easy to prove, generating marvel images from the marvel avengers movie can output frames almost identical to the movie, a topic that was trending on twitter a few months back, the culprit being midjourney, and now that adobe trained some of their data on midjourney, its clearely there's no "clean" database In fact, making a clean AI database makes it output trash because it simply doesnt have enough data so data laundering is the only option available for AI. And artists keep using and paying for adobe products are only proving how hypocrites they are, hating on AI while keep paying for the TRAINING process of it. Why not use Clip Studio? I remember how much hate Clip Studio got when they were trying to implement AI generation, don't see the same hate towards photoshop at all, AT ALL.

[@capuchinosofia4771]

Csp was trying to add ai generation? I didnt even find out! I thought the ai colouring tool was cool, but was there something else?

[@vogelstrauss9]

@@capuchinosofia4771 They wanted to make their own image generator like adobe did but after receiving a lot of backlash they scraped that idea.

[@colorfulbleeding]

CSP is the new tool in all /IC boards and for comics inkers now. Krita got one of the most powerfull AI pluggin there is and so is not ethical, Imo Rebelle 7 is the only other software where AI is banned for now .

[@3polygons]

@@colorfulbleeding Procreate's devs have taken a really strong stance (against AI) to defend artists, IMO, the most solid in doing so (I say this while I am not even an Apple/iPad user).

[@lapissea1190]

Adobe AI still uses the dataset from its old versions that are very much not ok. They just added their stock library in to the mix

[@BoroCG]

Is that so? I always thought they used their stock library and completely open CC stuff to make sure it's legally usable. Anyway, with the recent news about Midjourney images used for training it's all over for the cleanliness of Firefly AI

[@idkwayta1722]

@@BoroCG I thought AI images are public domain?

[@BoroCG]

AI images are who-the-hell-knows-what-to-do-with-this-crap domain

[@lapissea1190]

@@BoroCG About the first message, there has been artist using it and just putting in their name and would get something that is very much their style. Meaning, adobe is using a dataset that is abusing smaller artists same as basically every other popular image model. It has definitely gotten better though. You don't get random shutterstock watermarks anymore.

[@fx7105]

ai will become legally useable once we are all dead and all the works it was unconsensually trained on become public domain.

[@sc4r3crow28]

what pullover is that you are wearing?

[@BoroCG]

It's Cropp

[@ShyZShark]

So... Here's the thing, AI is currently, and will likely forever be caught in the cycle of copyright abuse. I'm not a fan of most AI, especially the way it is being developed and implemented, and I actively criticized you when you had said that digital art, or at least illustrative digital art such as paintings, are dead and that it seemed it was all about money for you to an extent. This is a reason why I had said that. Even if we were to assume there could be a collaborative effort of all these major companies (which will never happen, greed is too blinding), that still wouldn't overshadow the independent models that people are training. This, will always lead to tainted data, available online from those independent models, that can be imported into the more "legit" models, by people trying to take shortcuts to meet that next "proposed solution to problem x, let's have AI do it". I mean... the amount of corruption and corporate espionage that happened just to get modern tech where it is now, is rather insane. So why, if it has been done for decades to no serious fault, like the death of the major companies that committed those crimes, would it stop now? They've made billions off of the tech they stole from their competition, only to be fined in the millions over a decade later. This is just a repeat of that same practice, except now they are stealing from the general public in mass rather than just from their competition. Even if a lawsuit were filed and won, you literally couldn't compensate all of the victims who lost their works to one of these models. Corruption almost never becomes purified when both all the leaders in control of a topic are corrupt, and when you start with pure corruption, in this case, out of laziness to develop AI the fastest. And these companies will not stop. So, if I had to give a scenario where AI becomes truly legal, it would only be after copyright rule gets a major revision at a minimum and realistically, it would likely require the complete abolishment of it. That, or the courts would have to agree that AI is doing, what any other human is doing when it learns from a reference, which would produce a scary precedent about what AI could also do... So, no, I highly doubt AI will ever truly move out of the legal gray area. It started out completely tainted, and they aren't going to turn the clock back and restart for ethics. Even if we assume the core understanding of the tech becomes super good and you can train off of a super small set of pictures, say... 40-50 images, the copyright issues will still be a problem once you get into the weeds of it, because claims will arise that the core "learning advancements" were only possible by using a bunch of stolen work. So is the efficiency brought about by those stolen works, legally gray? Probably. Therefore any model produced with that, is also subject to the same issues. I'm not a professional on the topic or with the law, but that seems the most likely to me. For me, I complete any works I do, by building it from complete scratch, the old fashioned way before AI hit the scene. I don't want to deal with it, plus, for me, it just makes any of my works feel less real. Yeah, digital art is "less real" than physical to a minor extent, but a computer doing all the hard work, means I don't need to learn the fundamentals of my craft... Am I still an artist at that point? I don't want to be an extreme doomer or whatnot, and I understand where you're coming from when you want to use AI, as it truly does enable a single person to produce projects they couldn't have hoped to do by themselves otherwise, but that is the pipe dream of AI. Even I have that interest, but the reality is always going to be far less open and available, so I don't even bother messing with it.

[@SweetLitzLM]

only acceptable use of AI is if its just replacing using a stock photo. Instead of being in the adobe stock or any other stock image gallery wasting my time i would just generate one that did the exact same job. Using AI to generate art and pretend you are the artist is the problem here.

[@Vads0]

I don't have any problem with AI. It's just another tool

[@BoroCG]

The problem is, this tool is made possible by using artwork and photos made by professionals that never agreed to take part in this, and were never compensated for this. And while Adobe made a strong public effort to make sure things are done better, it seems like it's conceptually impossible to make it right due to the sheer amount of images used to train AI. And at the very least, this problem makes generative tools useless unless you want to risk losing the rights for your own work after using AI even for a small portion of it

[@grimsonforce7504]

Same here and I find the doomerism with AI hilarious. It's so bad I've seen so many people accuse the most basic art foundation of being AI. The kicker is the ones panicking are the same ones copying and pasting 3D models into their "art."

[@skully959]

pure wisdom

[@Emma-zr4tk]

I feel that AI is going to be both better and worse in the near future. Now it's only "good" bc its trained by actually good art made by artists or by real photos taken by cameras etc. The more AI floods the internet (wich we already see) the worse it's gonna get. Sadly I also think future generations of artists are going to be less skillfull than today

[@brylidan]

boro should put ya face on tumbnails. i almost didnt clic this video

[@Xomps]

Most people crying aren't even artist or have never made anything of actual value. They can't even understand what it is or how it could be useful, just hating for the sake of it. You may lose some subs for this but they are losing way more.

[@chrismoschlerdesignillustr5329]

To be fair most people praising it are also either not artists or are terrible artists, haha. There are two extremely loud groups of people arguing over this stuff while the people it affects the most attempt to quietly continue their trade while trying not to be driven insane.

[@ianalexanderreyes5890]

@@chrismoschlerdesignillustr5329 it doesnt affect anyone, that's on your head, or would you say that photography killed good old oil portraits?

[@tenneluna6948]

Where do you get that from? all of us who criticize AI are artists getting scrapped. Yet only non artists shill AI without any kind of criticism (what we call "AIbros"). So it's more the other way around, lol.

[@ianalexanderreyes5890]

@@tenneluna6948 i'm an artist and i rely very much on AI right now, life is easier, rendering is quicker, you are as biased as the other person

[@grimsonforce7504]

@@chrismoschlerdesignillustr5329 Quite the opposite the ones bitching about AI don't even have a basic grasp of art fundamentals. I've seen countless portfolios from smug self taught anti-AI artists. No concept of value, shape, rhythm, weight, perspective, etc I could go on. However as long as they receive praise from their instagram followers that's okay as long as they fit the narrative. AI isn't perfect by a long shot but I can with confidence I've seen more soul from a machine that I have with human artists. AI is a tool nothing more, if these artists want to be taken seriously they need to take criticism to heart and improve. Mediocre artists will just be that.

[@nestorlichev8531]

Normalizing AI art theft, especially as an actual artist is a disgrace. Those tools should be iilegal.

[@maikeru01]

It's the same thing as referencing or Photobashing.

[@nestorlichev8531]

@@maikeru01 no it's not. Concept art colleagues did a lot of photobashing years ago. They used paid subscribtion to websites from where they got their images. AI art is theft and nothing more. You cannot tell me that it shouldnt be illegal to sell images containing "make like Gerald Brom" and not pay the actual Gerald Brom from the money made.

[@marek_tarnawski]

​@@maikeru01 From what I heard in studios concept artists need to send the photos they use for photobashing to the legal team for approval. It is done so that they don't break any copyright. It needs to be from the right source. Using reference is usually looking at how something is made and then making your own design. AI mimicks data directly without understanding it which is very close to copying (and the perfect example is overfitting).

[@BoroCG]

Adobe pays artists whose images are used. I'm not normalizing AI though, I am saying it will probably never be legal

[@tenneluna6948]

Compared to his previous vid on the subject, I don't think this vid is normalizing AI, and i say that as someone anti-AI. It honestly makes the AI tools look really bad and cumbersome, even the beach pictures looked really bad and baby's first photoshop, the shore is a literal straight line.... you'd have an easier time photoediting it yourself or hiring someone to do it

[@electronicinfection]

Give this tech a couple more years. The improvements will be tremendous. AI tools are here to stay, like or hate it folks. Can you cover some of the new UE 5.4 features? Maybe the ones you are looking forward the most?

[@ivanmatveyev13]

Two more weeks?

[@marek_tarnawski]

I think we are reaching limits. I haven't seen much improvement for a while now. It stagnated and I read articles that AI companies are considering training on synthetic data because they are running out of images and text to train on. On top of that it's getting harder to get more clean data because internet got polluted with AI content. Also regulations are coming.

[@BoroCG]

It will probably take a few breakthroughs to go forward, hopefully something that will require a very small amount of images. That would make the legal/moral aspect of it much more controllable

[@marek_tarnawski]

@@BoroCG Let's assume though AI would be capable of being trained on only few images (like you use all of your personal art for that). In this case the model would probably be able to generate small variations of subject matter it already digested. I'm not sure that would be very usable. Maybe something like you train AI on your forest paintings and then you get a smart AI brush with which you can paint new forest using elements from your dataset. It's like stamp tool but adapts the forest shapes to the new context. A bit like you got geometry nodes in Blender and you can build things in parametric way.

[@ianalexanderreyes5890]

@@marek_tarnawski you can already create small AI models with your own art, it's called LoRA's and it's available with stable diffusion open source python programs, but it doesn't matter, you all wouldn't use it because it's still AI based, so unfortunate that artists decide to stuck once again in history with the traditional techniques.

[@gaymer5697]

Ai is just taking away printer artists jobs... If you are actually an artist then people care about your choice of composition and the story of the art piece, be it animation or an painting. Ai art helps creatives and hurst printers its as simple as that. So overall ai is very good for art. I dont care if furry porn printers lose their jobs. Also an "art style" has never been your own, people can make art with whatever style they want. If someone trully cares about your original "art style" they will buy your paintings just like people bought jackson pollocks splatter painting. that what collecting original art is.

[@onikaizer]

yeah, thank you for helping cleaning my feed from unnecessary subscriptions

[@BoroCG]

Unfair, you didn't even listen to me

[@KartikayMathur-y8e]

Adobe AI sucks What the hell I thought paying make it more Imperative but they are scamming you guys

[@BoroCG]

Well you don't actually pay any more for these AI features, you just keep paying the Creative Cloud subscription, and so far you get unlimited prompts and all, for this underdeveloped model. As I said though, does it even matter, since it's legally useless

[@KartikayMathur-y8e]

@@BoroCG True, Though AI is Absolutely amazing in terms of Inspiration, That's what I use it for atleast. Personality i realised that Opensource Models from Stable different and Civitai are more than enough to get any desired results. I would suggest you try it and play around. We might see a new game from you inspired by ai.

[@retrovisandomatt959]

sorry, gonna have to unsub forever

[@BoroCG]

I don't get it, I'm literally criticizing AI in this video?

[@retrovisandomatt959]

@BoroCG that wasnt the inital title of the video, and using the AI while criticizing it is a moot point, sorry

[@retrovisandomatt959]

dont feed the machine! * feeds the machine *

[@BoroCG]

Yeah, I changed the title to be more reflective of the content after seeing comments like yours. Why can't I have the conversation about AI? This is a baseless censorship that won't lead to any progress in this crisis.

[@retrovisandomatt959]

@BoroCG im not censoring you, though. its my personal opinion to just not follow you anymore, and criticizing it is good, just not while using it. the more you generate, the more it gets proficient at spitting results, which artists are fighting against