The Terrifying & Unethical World of AI Artwork (AI vs Artist)
ฝัง
- เผยแพร่เมื่อ 30 ก.ย. 2024
- The future of AI is a terrifying and potentially unethical world for many artists. And so today we've decided to pit AI vs Real life artists in a Dragon Ball Drawing competition! AI generated artwork has been taking the world by storm lately and I thought it would be both an entertaining and interesting idea to use our drawing video format to explore what we feel are the strengths and weaknesses of AI generated artwork! Join Totally Not Mark and his team of artists as they challenege multiple Artificial Intelligence models to a contest of who can draw Dragon Ball characters better!
MUSIC
Casualty (Karl Casey)
Red Ribbon Army Theme (DB)
Red Ribbon Army Motif (Path to Power)
SFII Animated Movie Intro (US ver.)
SFV Theme
Select Screen (SF 3rd Strike)
Ungodly (Karl Casey)
Sabbath I (Devilman)
Broly (DBFZ)
Funky Kong (DKC2)
Bardock's Theme (Cover by Android Orchestra)
Dudley's Theme (SF 3rd Strike)
Final Flash (DBZA Version)
Last Battle (Cover by Malcolm Robinson)
Rick's Backstory (Rick & Morty)
M1925 (DBZ)
MY PATREON: / totallynotmark
MY TWITTER: / totallynotmark
Edited by: Editor-San
Trev’s art skills are absolutely spectacular. I envy him.
true
Someone should definitely train a model on his scribbles, I agree.
he is the goat
Everyone goated
@@TheRealMagicMan he truly is
11:38 Okay but Editor-San trying to draw the hand multiple times, and then giving up, taking a photo and trying to use that, and still not managing to get it right, is the most relatable thing in this entire video
WAY TOO RELATABLE LOL
Artists be artists i am right.
AI struggles to draw hands, and apparently so does Editor-San.
Therefore, Editor-San is an AI confirmed.
we are all AIs
to be fair, he still did significantly better. the ai is one and done but a proper artist can keep trying till he gets it right.
Took you long enough to realize (I'm joking btw)
1 year later I rarely see messed up fingers these days
If we’re being real, no one likes to draw hands.
Glad to see that even the AI understands the struggles of drawing hands lol
Correct
@Txtthenumber_hut yo goofy ahh up
My art teacher couldn't too (easily that is) and he used to always roast me for my hands
it's because humanity has issues with hands in art, as a result, because it has more errors in drawing hands to model vs correct hands.
The issue is going to come when we start feeding objects into the machine where it can then properly handle standard photos and adapt from there into another style taking into task shading/drawing styles while mimicking real objects...
Scary bit? This has been around for 1 year.
1 Year and the AI art is this close... Seriously it will only get better, and rapidly so, to the point where it's going to be able to do what the gang here did... And that's not a great thing - it's why artists are trying to stop it.
Give it time...give it time.
As TF2 Soldier once said: *"Human grit always beats robot magic."*
Faxxxx
It depends on what it is. And how far ahead the AI is.
@unknown It's time to become metal gear, and improve the human race!
Me one day creating a true ai based on my own intellect laughing in the corner as my ai child beat people in art and math through my teachings
Important stuff!
I like how Mark took up the shintani type of aesthetic in his depictions of him and his team
yeah that was so good
It's not shintani style, he uses a lot of art style from namek saga or earlier and also a lot of yammumuro style
@@EnoYaka the colors and lineart just kinda look like that
@@EnoYaka The reference was to the animation in the beginning of the video lmao
As an artist I felt so bad for editor-san, he was struggling so hard with bulma's hand xD was such a weird angle to draw. Either way, great job, everyone, I prefered some of the coloring in the AI version( I mean, AI probably uses a lot of color theory because of the art it learns from), but the artists won this fight for sure.
What's an 'artist', grandpa?
I felt immediate pain with the constant erasing and retracing lmao. Such a mean angle
@@OrgusDin something a talentless Ai user (you) will never be
All that matters is the end product and he did pretty good.
Artist's will live on forever.
While It’s scary how quickly AI Art is improving, I think Human Artist will still hold on for a *Long* time. Human Artist will always have the flexibility and Quality that AI lacks. But time will tell.
It's an amazing TOOL for artists, not their doom. It's like how digital art was frowned upon by traditional artists until it was widely adopted. I can't for the life of me draw traditional but I can do pretty well digitally.
Its weird how you guys think AI art isnt just an extension of our own art.. This isnt a competition, AI is simply a tool that creates art using our own art as reference. Its a beautiful symbiotic relationship we should continue to grow together with.. Not oppose and fight against for superiority.
If it ever comes to that point, I'm hoping a meteor or climate change will wipe us out by then lol
@@CJKenjii ong
@@xavierlehew6746 By literally stealing Art on mass without consent, these AI gernerated pictures are even able to exist. Most artist have not given, nor ever intend to give consent. AI Art is Art theft. So no, the artist community does NOT want this to continue, it does NOT want this to grow. It KNOWS it is a danger to artist and mass copyright infringement.
So I’m a 3D artist and I’ve see recent AI attempt to generate 3D models and it brought me to a realization. Ai art in its current state and direction is an enormous waste of potential. AI could be trained to do stuff like topology, rigging, UV’s stuff that no one really likes to do but AI is arguably better suited for. AI could be used to enhance the existing art process instead of trying and imo failing to replace artists. A fast color pass, quick shading, turning you rough sketches into passable line work, stuff that yeah you’d still have to tweak and go over by hand but would definitely speed up the process. That doesn’t even get into all my main issues with Ai art buuut I just really think the current state of Ai art is really disappointing, it doesn’t feel like genuine progress and it’s a waste of a cool technology while simultaneously being incredibly disrespectful to artists
Edit: I just wanted to add something about good AI, there’s a tool that’s been developed recently that does a really good job coloring your animations, you just color the first frame and it does a pretty good job filling in the rest. Ai being used as an actual tool, training it off art it has permission to use and for specific tasks is a good thing and how I’d like to see this technology develop.
Hope it ends up like the meta verse
I've actually been saying this myself for a while now. AI is perfectly suited to help artists and animators with their current workload doing all the exhaustive background work.
See the thing is the people at the top of these ventures don't want to help artists. They want them replaced, because that way they can make WAY more money.
Are you and the people in this thread dumb? Like, legit dumb? Do you not realize we're working toward doing exactly that?
"AI art is wasted potential! Instead of doing this one cool thing, it should be doing this other thing!"
Like bro. Things can develop in a hundred paths at the same time. It's not mutually exclusive.
@@Klokinator Do you realise that currently these AIs are seeking to replace artists? And thus even if AIs that are solely there for making the tedious parts of art are made they won't really matter as artists will loose their livelihoods?
Don't call people dumb if your argument isn't well thought out.
All the art you guys did in this video is incredible. Truly striking and well composed images. The points you brought up are also very insightful
the title is very funny and hypocritical since both AI and Artist are drawing trademark and copyrighted material, they dont own. as well as any artist that does commission work i see all the time at comic con of artist selling fan drawings of Anime characters.
@@matthewjuarbe5826 Weak argument. Real artists don't need to use other images to create their own art. Also fanart is often just tolerated. AI art is copying other artists work and the producers claim it their own. It's so unethical.
In what world do you want to live? Where Big Tech super rich companies can get more rich by taking everthing all the artists of the world have put everything into it for nothing in return? A world were art itself may stop being human and everything gets more homogenous and lower in quality?
@@gez1009 no producer is claiming anything that's just speculation. everywhere i see people are showing Art generated using AI. also, also AI is taking nothing. also Rich companies do not profit from AI because they cannot copyright or trademark AI work unless they modify a significant portion of it. the law states AI Work is like public domain you can sell it for commercial profit but you can't copyright it which means anyone can use the images you generated in AI and create something of it. companies won't just switch completely to AI due to they can't protect the IP AI most likely will be used for generic basic things like brick stones trees grass generic environments to speed up development process which has already been done for decades prior with companies selling packs on Adobe davinci or any 3d rendering tool. i understand what youre saying but realistically Companies wont rely on AI alone due to they cant really profit off of it due to they cant stop another person or company from using AI images. due to they cant trademark or copyright it
@@matthewjuarbe5826 Many AI art producers claim it their own. Some even try to impersonate other artist. It is so easy. Just a view clicks. No, the winner will be greedy man with ties and big tech in general. They get the money. Because as you said. Currently ai art can’t be copyrighted. But the users have to pay for the service. So there is so much money already made. They earn it by basically selling the intellectual property of the whole world through a loophole. I respect your opinion, but I don‘t think you see the full picture. AI art is crazy impressive and if we don‘t push it to an ethical human oriented technology, we will regret it in some years.
@@gez1009 so how do you know the odels were trained using Copyrighted P which is against the law by the way i have used Midjourney leonardo AI and none of the models trained used any Copyrighted work you can actually look at the model yourself. the AI isnting on the internet looking at google images they are going off of what was fed to them. now if youre rawining you own model using copyrighted IP then thats another issue but the popular user friendly ones like midjourney and leonardo ai arent usign copyrighted IPs i checked asnd you can easily check yourself. also AI work can easily be detected
These are 100% my favorite videos on the channel, right next to the discussion vids. Seriously, I love all the fixing supers art, or 90’s redraw and everything else in those categories. I’m an aspiring artist and I love watching and taking in the techniques you guys show off and implementing something similar into my own works. Absolutely lovely things you do.
Same here
every drawing was phenomenal. the last 7 were miles better than the AI, and the first drawing is the only one that is somewhat comparable to the AI. great work you guys
Agreed, only caveat being I feel sometimes the AI wins in color compositing, simply because it uses official/profesional promo art as the data set
Due to a national geographic picture of a monkey that the monkey itself took, a lawsuit happened and it ended with the judge saying that in order for copyright to effect a thing, it must be made by a human. In other words, artists will be safe since there are always people who need the copyright to an art project. I mean, there's no way Disney is going to be fine with their Disney+ logo being public domain
That leaves out the small artists, which is the majority of artists. Those relying on commissions from people that would rather spend 10 bucks on Midjourney and get something passable than 50 on an artist and get something actually good.
Also not every company is as big as Disney or as scrutinized as Disney. So lots of small to medium companies will get away with it. Which is happening rn, seen a few contractors loose their gigs because of AI. And since most artists don't have the funds to sue the companies that employed them, these companies will get away with it. Also some artists will gladly take work by editing AI art. That includes lots of people from third world countries which are okay with lower wages and lots of 3D artists and other forms of artists who think they are safe.
Finally if/when Disney will think it's safe enough they will certainly lobby for AI art. They changed copyright laws before with the mickey mouse laws, what's one more law to change?
@@badreddinekasmi8919 in that case, I vote that any company that uses AI generated art should have their AI generated logos and art used for many unsavory things. I can't imagine any family friendly businesses would appreciate their "official logo" on adult websites. Nothing legally stops that. Or to use that logo for other businesses
@@Prime_Rabbit All we can hope for is regulation. Best case scenario would have been for this pandoras box to never be opened. But now is too late. So only regulation and adversary AIs will do.
The people who sued the photographer were members of PETA, and they called the monkey Naruto… yes, I am being 100% serious
Unethical yes. Obvious and vastly outperformed by actual people, also yes.
Then why are you seething, if you were even remotely capable of competing you would do that instead of just seething at it, but you aren't. Drawcucked forever.
@@OrgusDin Seething? Lol
I think you misunderstood. I don't use AI art. I was agreeing that human artists are better. No need to be unkind friend 😊
So, I will say that the AI is really good at visual effects in these pieces (the lighting and filter on the Broly one is crazy).
And the coloring. If AI just did the coloring I would use it
Yea they also have amazing shading and color choice. Just very iffy design choices 😅
One guy here said it perfectly. AI is best used for the boring effects, coloring, linework etc. The actual Artists draw the art and AI does the supplementary coloring, effects etc.
I honnestly like AI art as a base like you do in this video, it's fun to input a random idea and see a rough estimate of how it would look like and then correct it.
This is a good mentality to have. I think AI art is just there as a tool, not something to completely replace human artists.
@@brickbreaker3 a tool built on stolen work
@@brickbreaker3 I hate being human anyway, so i think it should replace human artists
I was thinking that the whole video, after all, what kickstarted these artistic renditions were those same AI drawings, correcting the issues and poses and creating something new. It surprised me when mark stated it couldn't be used as valuable inspiration
@@joaquinflores2741 There is litteraly a store in NY that make wedding dress based on AI Art and they are all beautifull, would love to remember the name.
I accept the challenge, feels like I'm competing with a real life Cell that has everyone's powers.
I'm not even halfway done with the video yet, and I can confirm that your entire team has already won.
It's not AI vs Artist really.... It's AI making a drawing and then artist redraws AI version with changes, basically stealing AI's drawing... True contest would be drawing from scratch and then compare. AI is better and will be even more better in 5-10 years just cause it creates x1000000 times faster with good enough quality which later can be enhanced with artists for their needs.
People keep shoving ai art in my face and they're foreshadowing it over real art. It sucks, it sucks, because no matter how cool it looks, it's never going to be better than the original. Someone gave me a Sonic the Hedgehog ai art, he literally had 2 noses, and he had his finger connected to his mouth. Like...what?
To be fair that AI was probably bad, they do struggle with hands, but with the right settings you can (albeit not necessarily the first try) get a really good result
The devs of midjourney and stable diffusion straight up said that “art is a tool and should be automated.” When told that art is a fundemental part of the human experience they replied with “oh its just a thing that were gonna outgrow and evolve out of”
Complete dismissal of the entire field and craft. You really think that they care about ethics? They are literal comic book villains given real world power. They have to be stopped
What about the devs of other programs?
Honestly after drawing the same backgrounds and characters for an animation so many times, id wish there was an AI that could do it for me, no fun in redrawing stuff that Ive already drawn
Look man, I like you and your channel but I can't believe you don't see the irony of saying "We've established that using it for any form of profit or acclaim is immoral" while literally profiting from its use through the monetization of this video. Also we haven't established it to be immoral, you kind of just make that claim without any clear reasoning.
look man, ur an ai bro and need to shut the hell up
And let alone that, the purpose of the video is somewhat thrown to the trash when they start tracing instead of drawing 😬
2:20 The best way to tell if something was made by an AI is to look at the hands.
TBF: you can generate until you get something that's fine, also always depends on the model being used. I've extensively played around with it, there's tools like inpainting that allow you to fix small mistakes the AI has made to get something just right(like say gripping something and it doesn't really grip/goes through/lacks the object). From my testing it still takes work, you might just be better off redrawing it yourself to fix it 😆
@@papabones8753 They need the hands thing to cope, you are pushing him off the ledge with this comment.
@@OrgusDin It's pretty funny really. I've seen so much anti-AI art content, it's mostly "hands" when in fact for me it's usually just 2 generations or some inpainting away.
@@papabones8753 The other massive 'artist' cope is that only text2img exists, not all of the myriad other tools that make them even more pathetic by comparison than they were when they were only comparing themselves with text2img. Holy shit I'm glad I'm not an 'artist' or else I would be absolutely required by some cosmic law to be a complete twat clown.
@@OrgusDin I could somewhat be called an artist...at least according to other people...I've put in my own art as well, works pretty well to improve it overall.
I've personally been most impressed by the aforementioned inpainting, it's pretty much magic to what it feels like.
I would imagine that the worry of stealing someone's art is only "okay" as it hasn't been tested. If disney found out your AI generated image has their stuff in it, they'd likely sue
They dont care as long as you arent hurting their bottom line or make too much money on it. If it just brings some views to your website or twitter, most corporations could literally give a shit lol. Its highly unethical though.
@@djentmas867 did they not sue a family for a child's gravestone having Mickey on it?
@Pokémaniac Guquiz if true, I think that's silly but they likely sued as they didn't want Mickey associated with a gravestone when Googled. If it never got buzz, likely never would have been an issue. But yea... weird thing to do. Clearly your company touched that family in a special way. Embrace that
th-cam.com/video/8eokIcRWzBo/w-d-xo.html&si=EnSIkaIECMiOmarE Good quick video on the subject
@@braydenb1581 ik Nintendo would definitely sue. "Don't profit off our product, we will sue anyone for any reason"
Why did you change the title like this man
11:00 lol editor-san supporting those "hands are hard to draw" facts
Honestly most of the limitations you mentioned aren't things that exist because they are not possible with AI its just they haven't been trained to do so yet/ don't have the feedback systems in place. You could definitely give AI 80s music and it would evolve something entire different out of it, if there is a positive feedback system like human listeners saying which songs they like from the AI this would speed up the process. Dont think that AI can not be creative, deep learning is modelled after how the brain works , uses neural networks that mimmick the function of neurons in our brains. Humans don't just come up with ideas out of nowhere they also have a series of inputs / inspiration and there is a random element to creativity as well. The AI doesn't just have to reproduce art in the same style it was trained on.
Im not exactly in favour of AI but as someone who studies computer science and does art the future is grim for artists. The AI is only going to get better at these things.
I agree, after all, I know also about computer science. Also, I hate when Humans think they are superior based on extremely vague ideas for which they often have wrong ideas. (in this case of creativity)
@@anderiko1816 Yeah humans need to be more humble. Assuming the AI is limited is why the recent developments snuck up on artists. It seems like a distant dystopian future but in reality we already have the tools and the knowledge it's just about execution and investment.
You have earned a like for the brief animatrix frame.
Is it scary and unethical that you buy things built on an assembly line? They replaced master craftsmen but apparently it isn’t a big issue because well they are just blue collar workers. Gotta love the outrage when automation starts creeping into the “creative” space. Lol
Small problem with your allegori. The outrage about assembly Lines stealing already happened. It is called the industrial revolution, where people protested the advanved machinery that took Jobs. Similar discussions has happened and different times as new technology has been introduced.
I think, intentionally or not, this video showcases the correct answer AI art has in relation to artists. Not as a tool to circumvent or replace real life artists, but rather as a tool to assist them. What the AI produced used as a sketch saved an immense amount of work for the artist by sheer virtue of coming up with a raw initial concept and a decent sketch to work over, through which artists can use to work over and add their personal elements and add the finer details.
Rather than rejecting the new tools of the modern era we can take a control of them and put them to use.
thats a good line of thinking. Problem is not just artist are using this tool. anyone is using this tool and taht means non artist who have no respect for creativity who will prepurpose and sell this stuff. Its doing more harm to artist than it is good right now. I think oneday we coan get to this point but ai art needs regulation
until it gets to the point where it gets good enough people go to the ai for art instead of actual artists
@@kossemoore3342 And why can't non-artists use them? It's not like all artists out there are altruistic saints who have no intent to profit of their talent. And it's not like non-artists don't have creative visions either.
Is it actually doing actual harm, though? I've heard a lot of suppositions, but do we have an actual impact? Have we seen the million dollar AI art NFT scams? Have artists commission incomes dropped or patron funds diminished.
I don't think it's "one day". I think we need to condition people in using new technology the right way right now that's it's newly released. Instead right now people are continuously instructing others on how to misuse it through fearmongering.
@@toasterowens8916 Even if it does, there are human traits that you cannot reproduce such as intent. The technical skill isn't the only thing that makes artwork good, it's a myriad of elements most of which cannot be replicated by a computer or a learning algorithm.
@@crushingon a good enough AI could replicate anything if given enough time
I saw recently that some judges passed some laws saying that any art form created from an AI WILL be fair use to anyone, and has no copyright, which helps a lot, but it’s still a big issue
It's based how you have no understanding of the legal system and also no understanding of AI at the same time, as well as no understanding of the reality of the situation and the fact open source has already cucked you completely. It's high level 'artist' cope and for that I applaud you, I never get tired of watching the bargaining stage of grief.
@@OrgusDin hoes mad?
@@blightcaster4388 You are, but you will get better, or not. It really doesn't matter, 'artists' are old hat and a stupid machine replaced you.
@@OrgusDin i am a hat now? damn i had no idea
@@blightcaster4388 Are you even an 'artist'? If no one has bothered to specifically train a model on your 'art' then the answer is no, so don't get too butthurt about your loss when it wasn't much of one to begin with.
I think this video really demonstrates a great use case for AI generated art. I've never been on either end of this interaction, but I would imagine for commissioned art, it may be easier for the customer to convey the image they want to create by using AI art, and a trained artist can use that as the jumping off point for an image that isn't lousy with all the errors a generated picture makes.
100%, I've already done this as a commissioner. I got about 5 or so roughly on model pictures from an AI then sent them off do an artist to finalize the design from. It worked perfectly.
I think people wouldn’t have such an adverse reaction to AI art if the AI enthusiasts online didn’t put someone’s finished product into the machine and then posted the new AI version under the original.
It's easier for the commissioner to make his own art and let the 'artist' get a real job if he has any talents besides being slower and less creative than a piece of extremely dumb technology that is still only a primitive shadow of what is to come. Crying online seems to be the #2 skill 'artists' have, try that.
@@OrgusDin I think there's more nuance to the arguments that is being missed in your comment there. Not to mention that implying commissioned artists don't have a "real job" is a very poor choice of words.
@@isaacfiresmith5250 you doing that makes you a scumbag.
If someone is commissioning human art ai in no instance should be in that equation.
A new android? Pretty sure that makes 11, nobody tell Trunks’ dad..
Interesting to note that I think that AI Bulma that Editor-San corrected looks like 18 more than she does Bulma. Editor-San did a great job turning it into a recognizable character
As an indie game dev, AI art is a massive help for concept art. I'm able to see a rough version of what my idea might look like and it helps me get even more ideas for the actual finished product.
I've talked to a lot of indie devs and one thing they've told me is that they wouldn't use ai art in final production as there's no legal precedent and ai art isn't really anywhere close to working back and fourth with a human artist, what's your take on it?
@@samuelkibunda6960 I 100% agree.
AI art should never be used in the final product because it could not only put you in legal trouble, but it could make your game feel soulless, since AI simply does not have the nuance and specifics that a real person has. However, it is a great tool to use in the pre-production stage of a game, and only then.
This topic usually is spoken through emotional lenses. Lots and I mean lots of misconceptions about AI art. The vast majority of AI generated images are totally new. Yes they were trained using images created by real artists, however that's not even an issue. Unless you point out to me some obscure artist that was born blind the argument is null.
The idea that each generated AI art needs to have reference to all pieces of art it used to generate the image is not only absurdly idiotic but also shows that people with this argument don't know anything about AI art.
About the creativity aspect. I wouldn't call AI Art and Ai Art, rather a generated image using AI. The prompt is always done by humans and not a machine. You still need to write it. AI art is still on it's baby steps and just recently learned how to walk. AI art doesn't know how to run or jump yet. Like a baby it mimics. When AI art wouldn't even need human intervention like today with prompts it is then when AI ART would have evolved.
However, this type of stuff is not going to happen soon. Not in this decade. Maybe 15 years from now we will see a new Model of AI which would be capable of this sort of stuff. I am not saying that we won't have significant improvements in the next 15 years as I am sure that even 5 years from now it will improve substantially. Heck even like 10-20 percent improvement would be quite significant.
As for the styles you could theoretically mix the styles in the prompt to create something more unique.
On ethics: If people accept comic book artists who trace or reference other pieces of art, there's nothing wrong with an AI using other pieces of art in its algorithm. Human creativity in and of itself is arguably just an act of employing previous information and combining it into something "new." No ideas or creations are truly original even in a human context. Pure creativity may not even exist in any objective sense.
Quite frankly, anyone who is complaining about the ethical implications of AI generated art is just uselessly virtue signaling.
Obviously, your art was better, though.
But art is the end result, not the process. If everyone could just transfer what they have in mind to the computer without any in between it would still be art.
In the futur we're probably just going to see people get paid for their imagination, then an ai generated image from it and if needed some ajustements by an "art technician" are whatever.
All the "artists" that only makes the "anime girls" stuff are 100% getting replaced by AI.
I used A.I. art to generate "game show contestants" for a powerpoint gameshow I made for fun. It worked okay however, I still had to take the A.I. images and adjust them to look a bit better. It took me about 30 min to correct the images and about another 30 min to find acceptable A.I. So, an hour of work vs my no drawing skill which would have taken me days was interesting to do.
Reminds me of one of the benefits I often see people bring up, getting backgrounds.
This video is probably the best discussion on AI art I’ve witnessed so far. You covered all the main pros, cons, short comings/flaws, and ethical dilemmas. Personally, I just see AI art as a additional tool for artists, hobbyists, and consumers to play around with, be it fun or concept. Regardless of what you think or feel, AI art is out there and there is no closing Pandora’s box. It’s unfortunately going to be abused for fame and profit no matter what, so I think it’s best that artists adapt and perhaps try to use AI tools to their advantage. That’s my two cents anyway.
yeah, AI will not replace artists, is just another tool in the toolset, people didnt stoped to draw and paint when cameras were invented
Human Artist: "AI artists cannot innovate."
AI artist: * gives Vegeta a unique pose *
Human Artist: * changes it back to the same pose Vegeta's been doing for 30 years. *
There is no creature more pathetic than the human 'artist.'
I never thought of the AI art controversy fully and why it’s a negative thing. Thanks for your insight and research on it TNM. Also, how long does it take you write a script for a video?
This was a fun video, and I dig the attempt to educate about the issues with AI art, but like many videos it doesn't explain and even misrepresents how copyright, intellectual property, and fair use actually works, and I think that's leading to misunderstandings about stuff like if the AI is "stealing" or the role of "creativity" in these debates. For context, I'm not an AI guy (in fact I do worry that corporations may undercut artists with it), but I am somebody invested in Intellectual property law, reform, and have followed cases & legislation for a decade. In particular, I'm worried that in a well intentioned bid to fight AI, artists may encourage legal changes that backfire and open human artists to liability too, not just limiting AI, as there's actually a fairly strong Fair use defense for a lot (but not all) of what AI's are doing, and any precedence set limiting Fair Use or expanding Copyright protection to stop AI might also apply and limit real artists too.
Obligatory disclaimer that IANAL and that my comment here is based on US law, for other countries it might be different: For those unaware, the main thing in fair use determination is if the derivative work is "transformative". This is a legal concept that basically is asking is the derivative work is putting enough of a new spin on the original in both/either a literal and abstract sense. This is an IMMENSELY complicated concept, though, and to be honest, it's sort of impossible to cover entirely within TH-cam's character limit: If something is transformative is just one aspect of many the courts consider when weighing a fair use defense: There's also Parody claims; there's the "4 pillars" (Purpose of a work, Nature of a work, how much of a work is used, and commercial impact), which a lot of sites act like are the only 4 but as i've explained there's other aspects too; and a whole host of other things. And Judges can and do invent new standards or weigh certain aspects more or less: You can satisfy a lot of requirements and still be infringing, or you can satisfy very little and still be fair use.... BUT, again, if something is transformative is *usually* the main factor, and this would be 12 pages long otherwise, so that's what i'm focusing on.
The video here almost frames AI as almost splicing together bits of existing images, but in reality no actual parts of input images are cut out and pasted together: It's gleaning artistic principals via examine the trends and patterns between the input images, which it uses to develop an algorithm that can apply the things it "learned" to generate new images. This has a VERY strong argument for being transformative: For starters, the actual medium is changing. If you write up a text description of a painting, that's obviously not infringement, and what the AI is doing is even LESS of a adaptation then that: it's merely documenting select principals of it's composition rather then describing the work in totality, and not just from it but thousands of other works, where any one of them is merely a small part of the new whole. If anything it's more akin to saying that a book on lighting and composition would be infringing on the images the author looked at while writing the book, even if the book doesn't actually include the images or even describe them.
If course, that's just discussing the AI algorithm itself, not the output images: I think that's more complex: As this entire video demonstrates, you CAN ask AI to generate images of copyrighted characters (like Goku), and that's obviously infringing on DB's copyright. You can also use AI to spruce up or "uncrop" existing images, which would likely be infringing on that image you had it modify: In both cases you can tell what the original character or image was. But if you're having the AI generate a brand new output image without specific copyrighted characters (say "Dragon in a swimming pool", then chances are no one image of a swimming pool or a dragon it used as an input is going to be obvious in the final piece, because it's just 1 of thousands of input images with little individual impact: It may not matter if the entire final image is derived from copyrighted inputs if only tiny bits of any one of them are still recognizable, but i'm admittedly not sure what the precedence is there with, say photobashing cases. I THINK it's just how much of any 1 is used based on what I've found, but I'm not sure. and there may be other factors.
In any case, something I AM sure is wrong that both this video and many other people bring up, that the notion that the AI lacking "human creativity" in contrast to an artist looking at references somehow makes it not fair use or transformative. This just isn't true, at least in current law, as far as I know: Certainly, putting a themetically interesting and creative spin on a work can AID a Transformative (or especially Parody) argument when claiming fair use, but it's not a requirement. You do need human input to GET copyright protection, see Naruto v. David Slater (yes, really), but that's a separate thing. Similarly, the fact that AI's don't need to learn and practice also isn't really important in if it's Transformative, and the idea that it can replicate people's art styles also doesn't matter, as Copyright doesn't protect "Style", just specific works or characters, things, etc.
This is where I think the risk with fighting AI via infringement claims starts: What if there IS a legal case or law passed changing that, and now creativity or requiring effort/practice IS a requirement in Fair use, or style IS now protected by copyright: Suddenly now you, as a actual, real human artist, might get sued because your art happens to stylistically look similar to somebody else's art. Or somebody making a piece of art that's a derivative work as a human artist now has a higher bar for claiming fair use because they have to prove it's sufficiently creative or high effort. Fair use is already WAY narrower then most people realize, and you of all people mark should know that. (If people are wondering why music AI generators only use royalty free music and care more about copyright, it's because the way copyright works with music is different: Music only has so many notes and the like vs the infinite spectrum of visual color, shapes, etc; so two people happening to make music with similar sections is common, and the case law with music infringement is ALREADY as such that in those cases, you're guilty of infringement often anyways even if it wasn't intentional. We do NOT want that happening with visual art copyright too. There's other ways music copyright is different too, like mandatory licensing)
And the same giant corporations people are worried about misuing AI to undercut artists (which, again, is a real concern) would LOOOOOOOOVE to have this happen: It would make it way, way easier for them to go after random people online with Copyright claims and C&D's: Imagine Disney Toei sueing somebody for making original characters that are just done in the Dragon Ball Style despite not actually being DB characters or using any copyrighted DB elements! In general, copyright expansions always benefit big corporations and hurt small creators: The corporations are too big to be sued for violations (such as when Family guy stole some guy's youtube clip then DMCA'd the original), and small creators are too small to fight back when the corporations sue them, even if what the creator did was fair use. Again, mark, you should know this.
Of course, it's possible that a court's ruling or a law in an AI case could be worded in such a way that precedence would apply JUST to AI (Again, beyond being transformative, say, the market impact is part of fair use determination, and AI due to it's ease of use and lack of practice/skill is extra disruptive to the art market), but the courts have CONSISTENTLY passed laws and made rulings that hurt smaller artists and benefit corporations: As I explained before, ultimately it doesn't matter how many Fair use principals you do or don't meet, sometimes you win or lose anyways: In practice, how rich and powerful you are often decides things, many times before it even gets to court since big corporations can force settlements or draw cases out and the other side has to give in, or they can lobby for laws to pass. As it applies here, we've already seen the Concept Art Association's anti-AI fundraiser state they want to get the Copyright Alliance industry lobby group Disney etc are a part of to be involved, who want to use anti-AI sentiments to expand copyright. The courts will be WAY more willing to listen to the big corporations at the lobbying table or filing Amicus Briefs then they will small artists.
So yeah, that's why I think framing what AI does as "stealing" or talking about how it's "not creative" or "can copy styles" is dangerous: It's well intentioned, but it's framing the issue and presenting arguments that, if they gain legal traction, could be disastrous to the rights of real human artists, as we can't guarantee that a court will rule or a law will be worded in such a way that those limitations would only apply to AI. As for how to combat it instead, I think there should be a push to get the AI companies to agree to stick to Royalty Free images on their own, like how the ESRB was formed without goverment regulation by the Games industry, and that people should just try to keep the precedence in place that AI images don't get their own copyright (which will discourage people from using AI in corporate contexts) without trying to claim the AI or the AI images are infringing.
i will quote you everywhere. Thanks.
Why did you write this much?
For one, the framing of this video isn't a legal one, but rather a moral one.
Second, the notion that ai art is not stealing from other art because "none of the piece is pasted into the original" doesn't hold any logical water. The ai model cannot output anything without its input art. It being obfuscated through the output does not change the fact that the artists image being imported is the only way that image could be output as it was.
@@TeeHallumsYT I get that Mark was mostly speaking ethically (I could have been clearer on that in my comment), but firstly, I don't think you can cleanly separate ethics from legality here, and secondly, the discourse around AI art has already gotten to the point where artists are rallying around legal fundraisers, the main one, again, has outright explicitly stated their intent to get megacorporation industry lobby groups involved and on board that Disney and co are a part of: I frankly think any analysis of the issue that doesn't acknowledge the legal risks here to be not approaching the issue responsibly.
I'm personally not that interested in debating ethics, because it's subjective and I think trying to come up with a plan of action that avoids the legal risks I outlined is more important then me trying to convince you my ethical standards are better then yours or visa versa. But to reply to what you said: If the actual images aren't being used and they're just being used to derive artistic principals and lessons to use to generate out of new images, what's the difference between that and artists using references?
The difference that mark and other people bring up is the "lack of creativity", but that seems to be a point being made out of a misunderstanding of fair use determination with legal case I mentioned. It's possible it's just an ethical standard they came up with without that context, but "creativity" isn't a ethical standard people seem to care about when making fanart or using references much elsewhere: I haven't seen artists claiming their work was stolen just because fanart or a referenced image isn't sufficiently creative, unless it's outright tracing and plagarism without credit nobody seems to mind, and AI art isn't doing anything as deririative as tracing.
The actual variable here that seperates what AI's are doing and human artists doing it is that of a larbor disruption issue: The idea that corporations will use AI to avoid hiring and paying artists, putting them out of a job. Which IMO is a very, very real concern. But I think that's what's problematic here: the use of existing images to learn from isn't.
The issue of AI and automation replacing human labor is also the BROADER issue with we're seeing not just with art but also text, music, code, and for decades physical labor and manufacturing: I think this is much, much bigger conversation then just art, and we need to rethink the idea that people have to work to meet ends meet to begin with in a world where machines and AI can do most of what society needs to run. If that was the case and people could pay for houses/apartments, food, electricity, internet, and other things without a competitive career, then would people still care about how derivative art AI is or isn't?
That's the real issue, IMO
@@MajoraZ The gofundme is a 200k cope and they cannot do anything about AI at all, it's open source for half a year, at best you are paying some 'artist' to go consoom on your behalf because they will never do fuck all else to stop the AI and neither can fossilized politician rats.
I will say that something that AI art is good for, like you mentioned, is brainstorming ideas for new drawings or even stuff like character designs. It could also be used for producing stuff like book or album covers if you feel so inclined (though you lose a lot of distinct identity doing that since AI art has little to no soul behind it).
My issues aside from the replacement of artists in certain industries like animation or what have you is that its also easy to pass off a lot of this art as real, not to mention you can feed an AI drawing an incomplete work and have it churn out something half decent out of that stolen piece of art. And its not like you can just put the prompt tags in the metadata because sites like Twitter remove the metadata from images for privacy reasons. And even if artists made a more concentrated effort of producing images of the WIP drawing and uploading those to Twitter, Newgrounds, Instagram, what have you to combat AI artists and show proof of authenticity, not only does that give AI artists more material to feed their software, but its not entirely impossible in the future for AI art software to come out that's capable of doing the same thing (because you can do about damn near anything with technology over a long enough span of time).
Simply put: I don't know if this technology should be legal.
No one needing human 'artists' anymore is the best part.
@@OrgusDin but can it make animation checkmate it's doesn't so shut up about it
why would it not be legal?
The advancement in technology like this is not inherently bad, and has the potential to be a ground breaking tool for even the people it's supposed to """replace."""
The really issue is that, like any tool, it can and *will* be abused by people who have the intention to do so. While I'm just having fun using prompts and seeing what comes out all this hubbub is necessary as to open up conversations about this technology as it goes forward.
@@OrgusDin No one needing you comment and bitching anymore is the best part
If AI can replace artists then it can replace writers and directors. The A in Art will stand for AI... that's not going to happen tho. I'm not worried anymore, I'm excited. It would be pointless to start building a new anime off AI designs, especially if I knew what I and others wanted, better than the computer did. After I've designed my main character, AI can save time drawing the in between frames and parts I never enjoyed. There'll always be a market for handcrafted work and the best projects will come from people know how to construct things by themselves.
I can control whatever I want, while being ahead of time and my input will always be necessary for the project to stay on point. Even if this means there's more competiton, there's no difference.... There are already tons of people out there who are extremely experienced and we can never catch up to all of that work. Infinite competition is just the same as infinite competiton. A starting drawing will always be needed to make a good project. A creative person will always be capable of creating a good design. If we know what we want, why wait for a computer to guess how it should be done?
Art doesn't need fixing
i could see the ai art being uses as a way to start drawing your own art... like you could legit feed some ai a photo prompt it to make a character then redraw it to create an original character
8:55 looks like Bardock clashed fists with C-Moon.
This video calmed me down. I was super nervous about AI art. Made me feel like my work isn’t special but hearing you speak about it made me realize I’m good
You are good. It’s just in the future AI art will be even better than you and it’ll only take seconds to produce. You will lose your art job to AI, first major field to be affected by AI
I guess I'm confused by the video as a whole... you say AI is illegally veiwing compiling images from the internet and attempting to create a picture in its its style. You're doing the same thing tho... your going as far as tracing the desirable elements from the AI image. If if what AI is doing is illegal. What your doing should be too.
The same concept applies to style, when people draw in other art style is that not theft of IP from the other artist? Moat of these people mad are the same ones who hated things like fivr of hold fan art contests, because it'll take money away from real "artist", or they won't be able to overcharge someone for their artwork. Of AI can replace you, then was your "job" really that valuable to begin with?
I think AI Art is a great tool for brainstorming but not a substitute for real artists, yet.
nice cope, very cool
I think the humans win on quality of “sensical correctness” (lol idk I made that up) as in, the art is actually making real looking humanoids, but honestly I like the AI’s lighting effects better. Like honestly even better than what’s in the movies and the shows. That’s just me tho.
Although that vegeta AI generated image looked pretty fantastic not going to lie.
Well, standalone images tend to look better than animation frames in general (see anime posters compared to actual anime).
that vegeta looked good at a first glance but look closely to it, look for example at the shoulders
8:41 WTF I JUST REALISED THE AI ART OF BROLY LOOKS LIKE HE BRUISED HIS NIPPLES
Art is a thing that we humans do simply because we want to, and we will always love to do it, whether being paid for it or not. People who say stuff like "let human artists become obsolete and replaced in the name of progress, like horse carriagers after cars came along" really have a shallow understanding of art. Art is not simply a job or product, it's not a chore! Machines should never truly replace what we love and want to do in life, what makes us human, the pleasures of life that make it worthy living. A machine is not gonna keep me from, for example, swimming if i find it fun! This is even truer for art: each person is unique, and we love to develop our own unmistakable style and vision!
There is so much detail that goes into a work of art besides a description and broad strokes that one person can feed into an AI program. If I just gave some basic prompt/description for AI art programs, and then did nothing with the resulting art, I would feel deeply ashamed if I said that I actually made this art. It's fun to play with, sure, but it doesn't make me an artist.
Maybe AI can be a valid artistic tool, but not like this! Animators use interpolation software to help their work in animation, but they are always constantly paying attention to everything the program does, and course-correcting anything that goes quirky. It's very different from people who just throw animation into an AI software, press render to interpolate, and hope for the best (Noodle's TH-cam channel has a good video on this).
Suppose that we can someday make an AI that is actually sentient and creative like us humans, feels emotion like us, has a unique identity, and so on, as we see in sci-fi. This AI would effectively be an individual person too, and it doesn't replace me any more than any new human being being born replaces me. Each person is unique. Such AI should be granted human rights.
Point is: maybe AI will replace humans for many card illustrations, and ads, stuff like that, which doesn't need to be truly unique or have a person's individual and unmistakable vision in the details, it doesn't have to special. But people will never stop wanting to create and make art, regardless if they are being paid or not. Art is not a chore, it's something that we actively enjoy doing. It's fun hard work. Art is part of what makes us human, that's why we'll never stop doing it. Not to mention that art is never just technique.
Ultimately, these are the reasons why I feel this video and many comments here are often sensationalist scare-mongering. I also recommend episode 43 of the 2000s Astro Boy anime. My mind always goes back to it whenever this scare-mongering happens. Please forgive the porn ads in the link.
www2.gogoanimes.fi/astro-boy-tetsuwan-atom-episode-43
I don't like to have this opinion, but from a glance, I actually like a lot of the AI art more. Obviously when you actually take your time to examine it it's a disaster, but again; at a glance, it feels more animated and flowy, while your guys' stuff feels more stiff.
Human artists will always be better than AI generated art.
Also, Mark, I have an idea for a video you should do sometime.
Every artist beat the AI handedly. I feel like it wasn't even close, especially when AI can't do hands, and create weird artifacts and issues.
The number of times I, and many others in skilled labor fields have been told to "learn to code" as industries become more and more automated has kind of made me numb to artists now facing the something similar. Welcome to our side of the aisle. It's all just progress, right?
Warning random story and a bigger post....
And people reading this, If I am factually wrong with something correct me, thanks.
Programmer / Systems Developer here. One of the main reasons I went into programming field was career councilors and parents told me back in middle school it was the safest job for my future, despite me not having any chance in the school type I went to, to ever get a job there. So was told, just do helper jobs and that I'm f***ed without focusing my whole life on it and do all school diplomas for the higher school types (wasted 5 years of my life going back to school).
And I did.... Despite me only having a passing interest in the field and would have rather learned knife / blade / blacksmithing since it's a nearly dead profession here, to keep the traditional art and culture alive.
If this was a good Idea I tell probably some random internet person in 10-15years, still am only 34 yo. ;).
Sorry about the unsolicited story, I just thought it would support your point nicely.
Back to the video now, I did play with AI art generation and think it's neat. It's only a tool and the fearmongering and outrage is insanity in my opinion.
The following is not only in regards to the new AI Art topic but can be generalized and I did.
People didn't want XYZ (insert random technology or invention which can be used for work related stuff) because (insert fearmongering or outrage rhetoric here).
Jobs change, technology happens, work environments and requirements change with the times. A professional Artist can create with AI help insanely beautiful Artworks in less time. So why not use it to your advantage?
The AI Art outrage people on the Internet would have died from dozens of aneurysms by now if they would work in tech. Some of us bi*ch and moan about the "trend chasing hipsters" (as I call them) too. But it is widely understood that even if you are not interested or do not like the new "trendy" things, it does not change the usefulness of the new developments for certain people. At this point I see our bi*ching more of a meme and good natured jab at it, since it goes both ways. To be fair I am one of the "old people" shouting at kids on my lawn ;). But would never think of raising drama about something so inconsequential oO.
These nonsensical appeals to emotion instead of weighting the pro's/con's rationally are really annoying me. And most importantly get the facts straight, do your own research, collect information from as unbiased and credible sources as possible and use your brain to formulate your own informed opinion on a topic. Why TF do I need to explain how to form their own opinions, instead of shouting in echo chambers.
To the point of AI art not being able to create art in a style without having seen it- I mean, I know I couldn’t draw in the style of Dragon Ball without watching it (or at least having seen images)
13:27 the comment in this section doesn't make much sense. Again, ethically there is no dilemma whatsoever because each and every one of the artists of this planet saw a lot of art, in many cases for free, before even considering becoming artists. Why is it bad for a program to do it now?
In addition, the concept of trial and error is basic in everything learning related, AIs need to do it and need all the possible support so that they go from the generation of random art pieces to the greatest possible specificity. Something that is not going to happen in a short period of time if this drama continues to be made about it. I understand that this threatens the work of hundreds of thousands of people, but it is not the first instance in the history of mankind that technology displaces the human workforce, it is inevitable to the extent that we advance technologically
Trev art is absolutely gorgeous!! Amazing art everyone!!!
I feel like calling this AI art is stupid. Because it's not actual artificial intelligence, it's just a computer program
Calling it a computer program is not very intelligent either, since in reality a computer program, no matter how advanced it can do, cannot do any of this.
It is much easier to call it AI and that way we don't get complicated.
After everything yall just went through...a video like this? idk fam. you are literally monetizing a video on an IP that you dint own...I noticed multiple people drawing in this video but...I only see Akira Toriyama. I'm just saying that you're damn near in the same position as the AI in a few ways, but honestly, The AI may have a better leg to stand on. Why? AI learns the same way do, by observing. How it learns is just about the argument artist have been making and in my mind, the answer is obvious. YOU HAVE BEEN FED DROVES OF IMAGES that you don't own or have the license to in order to learn to draw in this style (or in any style for that matter) as well my guy! lol it just took you 20 years to do it. TLDR: How much money did you make off of this video drawing Akira Toriyama characters? Did you pay for the liscense to learn how to draw like this? Is japan still coming after you footage that you dont own? using team fourstar for your comedy instead of your own? These questions and probably more next time ON DRAGON BALL Z! "Please support the official release!"
"AI Art" is pretty basic because its like a Baby that only learn what you make them learn. If the learning process is basic , what the AI will understand will be basic and the result even more.
Basically AI can only be as great as what we teach them
Taking the exemple of hands , we humans know what hands are , how they are attached to the body , how many fingers there are.
We have a thought procress that let us know when we need to draw 2 hands , finger position and more...
Ai dont because we didnt teach the AI how we humans think. And the Ai is not capable of thinking on its own , just do what we ask her to do .
Its possible to get the perfect result by chance because of how much creation the AI is capable of doing .
its only depend on the information the AI got .
When the learning process will change , the quality will improve
But in the end its only a matter on how much we humans are capable of understanding ourself and teaching it to the AI
It's fun watching an industry who thought they were going to always be safe from automation get automated.
All these artists complaining that a computer program used their art without consent while selling art of other people's characters and designs themselves.
Complaining that ai art is derivative is also silly because everything is derivative, you are taking things you have seen or heard and modifying it just a little, that's what ai does. And before you say ai can't create it can only copy you should really look in minute papers and all their ai videos.
I standby the idea that, once sufficiently developed, AI-Art can be used to streamline 2D animation.
If one can train an in-house AI off of a series of Key-Shots, one could then use (and subsequently tweak), the artwork generated thereafter for the In-between frames.
This is idealistic thinking, of course, but I like to imagine that a properly trained AI could be used to cut down on what is undoubtedly a grueling and thankless task (as much as I respect and prefer the art of in-betweeners).
Stop calling it AI Art. They are AI Images.
same thing
@@SuperScaryPuma nope, it's not art unless you put effort into it
@@Pikm4n Art is subjective, there is nothing that is objective for the criteria of art.
you can call it bad art, or good art, but it still art.
@@Joyscp999 the "art" in an art gallery that's litterally just a dot on a canvas is most certainly not art. Screw it, can I call the doo doo feces my dog smeared on the floor art?
@@Pikm4n then i guess you understand my point, if it is bad, it is probably not art, in your mind.
but if the AI does a good job at doing art...it is probably art.
Hand drawn will always look the best.
I think it’s clear that AI art is at a clear limit to what it can do, and all of your guys’ versions of DB art are amazing! Especially Mark’s Vegeta in the Takahashi style OOOOHHHHH MY LORD THAT ONE WAS GORGEOUS
lmao meanwhile 2 years ago it could barely sketch a cow, now you have to cope "bro look we totally won vs it" unironically, sad as fuck bro, cope and seethe
"I dont think ai generated images are worth taking inspiration from" mark says while making probably the best drawing so far inspired by the ai image :) well done mark that is a fantastic vegeta
But he didn't take inspiration from the AI image, he took inspiration from the DBZ animators, he just used it to fix the mess that was the AI image.
@@TwilightWolf032 he traced over it, literaly
@@pedrojustice Look at how much he changed sizes and proportions. He traced the first sketch.
@@TwilightWolf032 and if that's not taking inspiration, what is it?
Mark: Talking about the ethics of AI art.
Meanwhile Trev is making some fucking bangers of art.
I hope we get another episode of this, your artist videos are just awesome, especially as a wannabe artist myself. I do kind of agree with how it doesn't feel so right for AI to be using stolen art. The thing that worries me the most is how, even with all the jank, it's capable of doing like 75% of what an artist can do. And what especially I don't like, is it feels like it trivializes actual artwork done by people. People who don't understand the intricacies of drawing won't fully appreciate someone's work, when they're able to see tons of random drawings done by AI.
All 'artists' are wannabe artists minus the AI which actually is an artist.
@@OrgusDin nice bait
This was a great and reassuring video to those who might be scared of AI. While I don't like AI personally, I do see how it can be a useful tool. And I love the art you and your team created.
You realise AI art is generated by stealing bits from art that already exists, yeah?
@@godqueensadie Yes I do know that, and that's why I don't like it, because people can profit off of theft and unoriginality.
@@TitanXecutor I think AI should replace humans.
@@godqueensadie and?
The difference between AI and 'artists' is AI is a useful tool and 'artists' are just tools.
The failure in this video's argument is that it stands entirely on the premise that this technology is static. This is what it is, what it will ever be, and as good as it will ever be. That's false. That's like someone in 1994 saying "The internet isn't good for watching videos and movies. It's far too slow and low res". This statement became untrue quickly as technology improved. This moment isn't the limit of AI Art advancement; it's the very beginning. From this point on, it will only improve year after year. This video acts as if THIS is the best AI Art will ever be, putting the video on an incredibly short clock for becoming outdated. This will be a fun revisit in 5-10 years to see how far people were off. There are a lot of artists in denial over this tech.
Same old argument, same old answer. False comparison, just because something sounds right, doesn't mean it is. This video makes a point that AI needs, will always need, images stolen from others. Doesn't matter what speculation is made, this is fact. Any further argument made is based on that.
@@thinkbetter5286 And when people start using AI that was trained on their own art
One big problem in this whole AI art issue is clients who really couldn't care less how good or specific the art is, just so long as it attracts your attention.
Just look at the San Francisco ballet fiasco. They used AI art to promote their ballet. Even other artists throwing visual artists under the bus because they could get something good enough for cheap or free to promote their own thing.
yup!!
I'm an okay artist and I think the ai is perfect for helping set up a concept (and composition).. I don't think it'll take people's jobs but make em easier, at least until the ai gets better haha
5:01 was so corny
scared the shit outta me
I think you're wrong to say AI learns in a fundamentally different way than we do. The only reason why what humans do appears more creative and original is we are 'trained' on more than just other's art. We're trained on our everyday experiences, which is just vastly more data than ai gets and unique for every person.
'Artists' love misrepresenting how the AI works and half of it is because they don't have the capacity to understand anything but scribbling on paper and half of it is because they are intellectually dishonest grifters who know the well of commission money has run dry and they will need to find unskilled labor in some shoe factory somewhere to eat (the shoes.)
Nope he is right. AI I models as of now work on probability operations not decision making. Case in point you can make chat gpt to agree to do anything by just forcing it cause it doesn't understand the context. Also if you tell a child to draw a horse and a AI to generate an image of a horse. The AI will have the face right and stuff but miss no of legs where as a child will have legs cause children have an abstract concepts of what an horse is that is not just pixels. And a child can recognize if that the horse has it's leg blocked by something. Where as Macine learning algorithms can not. Also of a horse has a birth defect human mind can understand In a single example that there are exceptions to the 4 leg horse rule.
This is just a blatantly wrong assessment. Unless you’re drawing/painting photorealism, the most important aspect of human learning is stylizing and creating shorthand for real anatomy to convey ideas without fully replicating them. AI doesn’t do this. It compiles other works to create Frankenstein products. Nothing an AI creates is original in anyway. There’s no distinct choices being made to deviate from sources because AI cannot deviate.
@@Watermelon_Man Ai does not compile other works and combines them. Ai learns how to detect patterns and creates a math formula to map words with pixels. When it gets a prompt to draw a horse, it's not looking through it's training data, it doesn't have ANY images stored in it's memory to reference. All it has a math formula that's been fine tuned to convert the word horse into an image. The nodes in it's neural network have learned to represent abstract shapes like curves, straight lines, heads, legs, colors, etc, and it learns that horse is specific a combination of these various smaller things.
As I said before, Ai only seems unoriginal because of how limited it's data set is. While an ai might be trained on 1 billion images, a human is streaming live images every single waking moment of it's life into it's brain and also has a subjective experience of each image, smell, sound, has context to everything it's experiencing, all that trains you to have a much MUCH larger range of ways of expressing an idea. But ultimately, everything you create is derivative of something you've experienced in the past.
Not a fair contest because you had a final image of an AI before the drawing and therefore you only had to build on it. To make it fair you should have drawn an image using only keywords the AI used.
AI art, and how it's regulated now and going forward will cause a hell of a lot more problems, and it's already leading to many questioning wether actual artists will be able to carry on as their output is outpaced from the sheer efficiency of it, even though in many cases it's beyond terrible, and even if most of it was decent through good, the push away from actual artists is not worth it at all.
Human 'artists' can't compete, time to cope.
@@OrgusDin In your case, time to piss off.
I'm actually worried that AI is capable of making art because I don't know how many more fields AI can excel at and potentially replace human workers. Yes, AI just regurgitates information it's been given, but isn't that basically what people do? The difference is that AI can absorb _far_ more data so AI will catch up to human artists very quickly.
However, AI (as it is now) is hard-capped by humans. Since it needs our input to draw from as inspiration, eventually it will catch up to us and have nothing more to copy... Unless it starts copying other AI-created work and I don't think that would look great.
Trunks tried to warn us
not really
11:57 Dude, that ain't Bulma. Those are characters from Dragon Quest IX.
Yes, the point was Editor-san was referencing modern Toriyama work lol
@@AnimeAjay Ah, okay. I'm a huge Dragon Quest fan, so it just immediately jumped out at me, lol.
I like this video, try promt vs AI. Like the artist don't see what they are competing against.
Like "Vegeta in a kimono using Kamehama in a iceland" and you all draw the same promt and then feed that into an AI generator and compare the best against your pieces.
Mark = Fail
Ajay=Fail
Trev = Win
Trev = Win
Editor = Win
Mark = Win
Trev = win
love the Ryo-Mark at 0:10. could this mark a Devilman 1972 manga review oncoming?!! i can only hope. the Devilman Crybaby OST used once in the video gives me joy too.
Ai art either makes an inferior version or something completely different
We haven't established that using it for any kind of profit or claim is immoral lmao
I enjoyed the drawings but the 20 minute Concerned Artists Against AI Art desperate PSA is cringey. If your artistic value is being threatened by low quality art then you were producing low quality art.
Great video, I think y'all did better than the AI in all of these. As for my thoughts on AI art itself, you bring up some good points, and I understand where you're coming from, but AI is already here, and there's nothing we can do to change that. So unless we're dealing with legally or morally questionable situations involving AI, there's no point in fighting it. AI as a whole looks like it's only going to get bigger as time goes on, never mind specifically AI art. So I think our best bet as a species, is to try and figure out how to use all AI technology as a tool, including AI art, while also not using it like a crutch, so the AI doesn't get powerful enough to render humanity obsolete, while also making sure we don't allow bad actors to use AI technology for malicious purposes. If we want the best possible outcome from this scenario, which is a future where Humanity doesn't succumb to AI, instead not only surviving, but thriving off it. We're going to have to play a very delicate balancing game with this technology, and I know that's going to be tough, but we as a species are known for doing extraordinary things, and I really do think with enough planning, caution, experimentation, teamwork, and diligence. We can pull this off, It's difficult but not impossible.
Editor san STRUGGLED with those hands but WON
What software do you use for your drawings? It looks super clean and I am trying to get into digital art instead of just paper and pencil
I can't stand the fact that some artist have been perfecting their craft for decades just to be superseded by some overtouted washing machines
(Oh god I'm so sorry, I just looked up here and realized how much I wrote jfc I can't believe youtube is even letting my post this book of a comment)
AI researcher here,
I wanted to provide my own perspective on things, as well as give people a bit of a deeper understanding of how these models work.
First off though, I do want to say that I agree wholeheartedly that these tools shouldn't be used for profit, and at a fundamental level it is important to make the distinction between open source models where all the data, code, and trained models are provided for free for anyone to use or modify, and closed source ecosystems where you are beholden to a company (likely through a subscription) for access and usage. OpenAI created dalle and dalle2, but their models and source code for projects generally aren't open at all, while stable diffusion is entirely free and open source.
Even on the corporate side, I think that its important to note that the research teams working on these systems weren't necessarily starting out with the intention to create something to replace artists, however, regardless of how it is approached, this will happen. It doesn't even matter if it becomes illegal to use such programs, as they are faster and cheaper to the point that it will cost the likes of disney nothing in legal fees to keep using them and lay off artists.
While its not impractical to say that in many applications of art its likely that artists will learn to use AIart as another powerful tool in their toolkit, there is no sugercoating the fact that this tech is as revolutionary and breaking as the innovation of recording music was for live performances.
On the open source side, stable diffusion (along with other research groups) generally started with the stated goal of reproducing the work of companies like openai and google, and providing that for free.
Some people have discussed various ways to create laws to limit AI or datasets and protect creators through copyright, but I am incredibly wary of these, as it often aligns incentives with the media corporations who would be all to happy to further limit fair use and profit off of fan content.
I also think it is more up for debate whether these models are capable of something resembling creativity, and the arguments for lie in the specifics of how these models actually work, which is often communicated in heavily simplified and misleading phrasing, so I'd like to take a stab at clarifying some general specifics of how these models work.
The key aspect of many of the current state of the art image generation AI is the concept of diffusion, which starts off with a bunch of random noise, and then iteratively attempts to approach a stated goal. For image generation AI, this means it starts with an image of random noise, and then takes the given prompt and makes a pass at the noise, slightly changing it in order to find structure out of the randomness that fits the prompt. This is quite innovative, as it means from the same prompt, random noise can create completely different generated images with the same intent, and in training there are many different 'right' answers, allowing the only a few gigabyte file-size of the model to have learned an impressive amount of general knowledge.
I think an important thing to note is that, once trained, the model is static, and unless it is fine-tuned further down the line, it doesn't learn or see anything new. Once it is trained, it doesn't have any access to the datasets anymore, nor does it search the internet, everything it 'understands' (in this case about art), is baked into the weights and biases of the neural network(s) its comprised of. The scale of compression here is unbelievably immense. Stable Diffusion 2.1 is about 5GB on my hard drive. It was trained, not on a few thousand images, nor a few million, but 5 BILLION images scraped from all around the internet. The LAION5B dataset, is open source, and anyone can search it if you would like to see if it has your art in it (it probably does). To put that scale in perspective, that is more images than seconds you have been alive, in fact, given current human lifespans, it is more images than seconds in the longest living human to date, and in total takes up ~250 TB of hard drive space.
In the case of stable diffusion, saved as float16, the model itself is only 2.5Gb. Because of this insane 'compression' and the from noise generation process, in practice it is actually very difficult for the model to recreate specific images from the dataset, except in specific cases where there are many versions or references of the same art pieces, such as something like 'Starry Night', or the Mona Lisa. To a lesser extent, this is how these models understand what specific characters from different works of fiction look like, which is also why creating specific characters tend to actually be quite a pitfall of these models. It may get, to some extent, who some of the characters in DragonBall are, but its likely using the similar styling to fit them together very tightly, if that makes sense.
Anyway, long winded way to say that these models do not search the internet for images, everything they understand is encoded into them during training, and that encoding is DENSE. It also doesn't do any sort of classification or narrowing down of the dataset as it generates a new image. This seems counterintuitive, because one might think that if you ask it for art of a specific anime character, that it should be pulling up references of that anime style, and of that character, in order to generate a new image. But due to the nature of how neural networks work, its querying its entire corpus of knowledge with every pass. One might be able to say that there is a degree of self-organization in terms of how things are encoded into the latent space, but at this point I'm throwing technical jargon around as if it answers questions, which it doesn't.
The 'creative' advantages to this, is that we can do things with the AI that aren't strictly possible for a human to do. For instance, we can ask it to create images that travel between two prompts, and we could see what the halfway point is between two styles, two characters, or what the average of completely different concepts look like, and it will do it. You bring up that many genres of music and styles of art of born from and developed off of one another, and I think that is important in this context. Given the right prompting, these AI can make those same stylistic developments, without being told what its aiming for, and through this I believe its entirely plausible that it is capable of genuine creativity, but its also important to know that it if it is, it learned that from us.
I'm not really sure where I'm ultimately going here or if I had any central point to these ramblings, its just been a crazy few months in the AI research world, and I don't think anyone thought we would be here now. I would have put this kind of AI a decade or more away just a year ago, and I don't think any of us fully understand the impact its going to have on every industry, especially the creative arts.
Also, fun fact, this terribly incoherent assortment of thoughts was once even more terribly incoherent, and then I threw it all at chatGPT and asked it to help me organize and refine it a tiny bit. ChatGPT is also quite capable of programming, moreso than dare I say a not discountable amount of college compsci students, and its also capable of writing comments, summaries, and taking feedback, which at least puts it ahead of me. While I do fear somewhat of its ability to supplant people's jobs, even my own, I also enjoy bouncing research ideas off of it, and asking to code things I'm too lazy to write up. What I don't like about it is that the model is entirely closed source. OpenAI has chosen to make it free for use now, but that may change in the future, and the true path forward, in my mind, lies in open source competition providing these tools to everyone for free. I believe that this sort of code generation AI, in the not to distant future, is what will empower small open source teams to compete head on with walled gardens of the likes of Adobe and Apple, and I think that open source and collective forms of automation and AI are the only way to stop the worst parts of capitalism from continuing to ruin humanity.
And sure, AI has lots of potential for downsides, and as these models get exponentially more powerful into the future we need to figure out the right and ethical ways of interacting with them, but at least they can generate cute pictures of cosmic cats of various festive varieties.
Okay, reading this through again (which I probably should have done before posting it,
how the heck did I write so much and also still feel like there are 50 million things and angles that I didn't cover or bring up for discussion aAAAAAaagagghh
A kind thank you and the most delicious imaginary cookie I can summon to whoever reads through the whole thing
Great thoughts and ideas, this has been my approach to thinking around a lot of the moral and legal challenges brought up by artists. I hope that we can all find ways to use these tools to accelerate the creativity of the human artists using AI tools rather than being adversarial.
Once again thank you for your thoughts!
this is gonna be REAL weird to say, but i think the ai's art has more emotion to it. while i very much understand that the ai art is riddled with illogical and impossible designs, its lack of specificity really focuses on the motion (and emotion) of the piece. with the increased detail, some of the associated emotion of the piece is lost.
except the last two pieces, the human designs were clearly superior in all regard.
Thanks for this video, too many people demonizing AI
I really relate to editor-san having to redraw that damn hand angle 10 times 10 different ways to get it satisfactory 😂
Great video Mark. You guys do amazing work.
I agree with a lot of your points, including the ethics of where the AI sources it's inputs. However, as a person with no art skills whatsoever, AI art has been a great tool for me. Often times I require images for games nights, table top roleplaying games (like DnD), and for casual hobbies. In these cases I often need something quick, specific, and doesn't have to be perfect whatsoever. If I have an NPC interaction, a one off character, or need to show a landscape, AI art has produced some awesome stuff. While I still use commissions for player characters and important people, the rest is certainly not something I could afford (or is worth) to commission an artist for every single time.
I've commissioned a lot of work off of various places (on the internet and in person). While I've had some good interactions and found some artists I can trust, I can also say it's not all been great experiences. I've had many cases where artists would be misleading regarding the quality of work (posting examples not representative of their delivered product). I've also had many cases where the artist I'm working with would not communicate at all during the process, miss agreed upon delivery dates by weeks, and not be willing to do requested revisions (even though it was in the terms of agreement).
I don't believe AI art can replace artists. However, I think it can be a really great tool, especially for those of us without artistic talent and years of honing their craft.
Please be careful with the naming of videos, there are a lot of missinformation and people who would read the title unironically.
Edit: Oh.... Well.... see the bellow essay:
Putting aside the quick implication of copyright violation, because this is frankly one of the most clear cut fair use cases I can imagine, (taking a composition and changing the style, or taking the style and changing composition, is perfectly allowed for any public work) the idea that an AI taking inspiration is inherently less ethical than a human doing the same is….problematic.
Partly, because neural networks are by definition made to imitate brains, the process is in fact basically the same, the only difference is that us humans have way more experience in seeing things from a ton of angles. One can’t really criticize the AIs method for not being creative without running into the philosophical dilemma of us humans being the same. If we could imagine completely novel visuals then imagining a new color or a 10d landscape shouldn’t be too hard, right?
Secondly, these AIs aren't acting on their own; they're another tool in the still human effort of trying to recreate things. It follows then, that one could not restrict it in a way you couldn’t restrict humans. The only reason to restrict its sampling then would be….Idk, it’s making it too easy? That just sounds either defensive or elitist. Maybe we shouldn’t have digital art? Doesn’t not having to buy supplies lessen the commitment necessary, and thus artistic merit of a piece?
I’m not saying that this is your argument, but it does follow the same logic as the middle of my last paragraph, which you might on some level believe. The difference lies in time commitment versus monetary commitment, unless we see the AIs alternative as commissioning, in which case they are both monetary.
ps
13:18 The numbers here are just…..No. (At least for any of the household names)
Less go I don’t got to pay somebody to draw me something 🗣️🗣️🗣️🗣️🗣️💯💯
ai artwork is like someone breaking the law of equivalent exchange
The laws directly don't matter to us.
14:22 great video but what is the point of the AI art if you're just gonna do whatever you want? (Angle and pose)
Why do artists care about ai replacing them if ai art is bad in their eyes?
Because they have bills and families too feed, not to mention their work being stolen and used without consent or did that just fly over your head?
@@CJKenjii but they’re worse. So like, no one would use AI over an actual artist. Also I thought art was about self-expression? Why do artists care about bills and money?
@@robertwyatt3912 Geeee I wonder why they care...
@@CJKenjii is it because they were actually lying about all that self expression stuff
@@CJKenjii Well they better go find a way to pay them since their grift is over lol.