What would good AI art look like?

My Patreon: / cuck
My Twitter: / philosophycuck

[@WannabeMarysue]

"Name One Thing In This Photo" is a viral image made by AI, which depicts what appears to be a messy bedroom, but on closer inspection nothing is identifiable. Its one of my faves.

[@f.t633]

just looked it up. it feels incredibly creepy

[@Longuncattr]

I still remember the feeling I had when I first saw it. Completely lost and like I could go crazy. Hell of a thing.

[@AKK5I]

Was it actually generated by AI though? It says it was posted first in April 2019, before any mainstream AI image generating tools had public availability.

[@lesale3024]

It perfectly replicates what I feel from dissociation sometimes.

[@f.t633]

@@AKK5I oh this is interesting 😯

[@WendingWind]

Tendies123 is literally the most generic internet username, yet so memorable.

[@PyReDZN]

Tendies123 is a real one too, I see their name in the thank you section for like every video this channel has posted it seems

[@azaraniichan]

I always listen through the credits just to hear that name, it's a sad day

[@andrewprahst2529]

Tendies123 Tendies you and me. Tendies, have some fun Tendies321

[@mrosskne]

TENDIES123 WILL RETURN

[@PoolNoodleGundam]

The point about how photography made specific use of existing and novel techniques to create art in the new medium is sticking with me on this one. I mod games, and I use AI generated images as placeholders (anything that makes it to the nexus is always manmade) while I'm building the UI or playtesting, and I've realized that the thing about AI content that rubs me the wrong way is the complete lack of creative influence over the process. There's an art to constructing the program or constructing a prompt, but you never get to feel the creative interactions that, to use the example example, photography has. You can't adjust your angle in the moment, you can't ask the subject to lean in more, you can't spin your lens. It doesn't think, it has no ability to conceptualize, it cannot feel. You can only ask it to try reinterpreting the image set, to have another go at reorganizing someone else's work.

[@WannabeMarysue]

But you DO have input. You can use more AI techniques like inpainting, which can be used to dramatically repose your subject. Or switch to a digital editing program like photoshop for their suite of image filters and tools.

[@spectercd4357]

Also it is up to the artist to do the thinking, conseptualization, and feeling so they can guide the AI to the right direction.

[@spectercd4357]

@@WannabeMarysue Also another way for this is to edit the prompt while keeping the same seed. This way you can see what effect each new word has, and you can change stuff like the mentioned angle, pose, lighting, and colors without altering the general composition of the final image too much.

[@MantasticHams]

There are plenty of ways to intervene in the process already, from before the generation to the very end of the process, it would be laborious to list them. BUT, there is a level to which i agree, though i think as things progress we will see a kind of multi-modal development in the open source generators, where the unique elements from things like DragGan, where you can drag elements around using various gui functions and turn someones face, widen their stance, open their mouth, and raise their arm, will somehow be integrated into the more overall high quality stable diffusion, which also has other unique elements of control over the image through controlnets and other plugins. Combine all that with a better understanding of natural language by havingf a more direct and collaborative connection with text AI like ChatGPT, and perhaps a more thorough understanding of space from models trained on 3d models, Photogrammetry, and NeRFs. This is all hypothetical, but it seems like the natural direction eventually, there are already rudimentary steps in that direction but the issue is it is very surface level communication, as if chatgpt is jsut typing something into the box for you, instead of intimately understanding the image generators various layers and injecting input that a human couldnt rationally input by hand, based on natural language (or other inputs), to generate the desired result.

[@salmadys]

What you are describing is the lack of understanding for visual language. Prompters don't need to understand color and light. They don't know what looks like bad Contrast or cluttered shapes. they don't engage with nature like a photographer does. So The work of prompters has an air of alienation. Like they are talking on borrowed words to say nothing, and pretend that what they are showing has meaning.

[@arms7260]

tendies123 😔

[@misanthrope123]

End of an Era

[@MeowImages]

I'm a woman looking at a ton of AI art, and I remind myself that the people typing prompts like "Extremely beautiful Pantene hair model, white conditioner dripping from her hair and mouth" are just a small minority of society, and that the AI-generated pictures from simple prompts (that don't overtly steal from artists, via "in the style of...") are just a milquetoast amalgamate of Instagram, Deviantart, etc. But it's still weirdly depressing. It's like holding a mirror up to society, and it reflects back on us. Maybe that's what makes some AI art so ugly and upsetting.

[@3xsxs953]

It reflects that most people lack taste and imagination, which is certainly depressing in a sense.

[@nadia-v]

the hair conditioner coomer prompt fucking SENT ME :D

[@leovalenzuela8368]

Holy shit that made me profoundly sad. You’re right and that sucks.

[@priapulida]

@@3xsxs953 also time and/or willingness to invest time, prompt engineering is more artistic skill than engineering, and one simply needs time to develop taste and imagination too

[@McDonaldsCalifornia]

White conditioner 🥶🥶

[@plateoshrimp9685]

Here's to Tendies123. Hope they're doing okay!

[@augustlovesjosh]

I saw someone talking about their company hiring AI prompters instead of illustrators and they were unable to make small specific changes asked of them. currently, most jobs that artists do would require them to be able to do this.

[@nuance9000]

Sounds like the company hired some consultants 😅😮

[@TheEvilmonkey25]

Sadly I think the most effective thing ai art has done was demoting illustrators, so that instead of being a creator we will now be cleaning up art made by AI (which requeries as much effort and competency as creating an illustration from scratch). Salaries will be slashed throughout the industry.

[@MaakaSakuranbo]

Well you'd need inpainting, not just prompting for that. But yes, that's one of the current drawbacks of AI

[@blueenvy010]

One interesting piece of AI art was a project that tried to generate images and their "semantic opposites" so to speak. They used the textual embedding of the prompt, (a vector in a high dimensional space) and the inverse of that embedding vector to generate two images that have the maximal semantic distance between them. As an example, through this process one would get a boring image of "a pretty women in a red dress in a corn field" and what the internet at large considers the opposite of that which I found very interesting and kind of revealing. Although relatively simple I think this is a small example of the kind of interesting AI art that can be generated through the unique characteristics of the medium. I imagine one could come up with a bunch of cool art projects that could utilize the latent space of such models, training different models on different data, remixing those models, intentionally collapsing the models by training them repeatedly on their own output and so. As a computer science PhD student it makes me sad that most AI art is this uninspired, boring, low effort and low quality schlock.

[@vfanon]

Can you link to the piece in question? I'm interested.

[@iankrasnow5383]

Of course it's uninspired, boring and low quality schlock. Making a text prompt is so easy anyone can do it. It takes no training or taste. Even without skill, people can make things we might recognize as art just due to creativity and vision. But it takes almost no creativity or vision to feed a text prompt to Midjourney. It's all uninspired because it's so much easier for people with no interest in art to very quickly create crap that looks cool, so there will be a lot more of it. But artists who care about aesthetics and creativity are already able to use AI tools to make interesting work. Btw, do you know where I can find this project that outputs the inverse of an image's vector? It sounds interesting.

[@mtdfs5147]

​​@@iankrasnow5383 This is what I've been saying. If artists embrace this tool that is a godsend to the entirety of the creative process they might actually have fun and be able to keep up with the rapid advances in AI. It's a tool first and foremost, if they can use it in some way to make their art better or faster why not leverage that?

[@blueenvy010]

@@iankrasnow5383 I think I saw it in a Reddit post titled "A bizarre experiment with negative prompts"

[@mrosskne]

so no examples?

[@TheCanvasArtHistory]

I love how you tackled the whole debate of "can AI-generated images be art?". By framing it as you did, you made the point (better than I have) that, just as not every picture is a work of art, not every AI-generated image is art. I completely agree and I have taken the position that "Ai art isn't art" simply because there hasn't been that much "good AI art". Thinking about what makes good AI art though pushes the discussion and the reflexion a bit further, which is greatly appreciated! Great video as usual!!

[@alexanderthedude5474]

damn didn’t know you were based lol

[@generalfishcake]

No AI picture is art, because it's stolen without consent from other artists. AI art is not like photography, it's like stitching together works from others & claiming it as yours.

[@s1nd3rr0z3]

​@@generalfishcake"Stitching together works from others" you mean like a collage? A thing generally recognized as art? Fuck your intellectual property who cares about "theft" of IP

[@Michelle_Wellbeck]

your comment is so generic an AI could have wrote it

[@snowdevil002]

i don't consider your emotional appeal video even on the appropriate level to have a discussion in good faith about this topic.

[@itsemmallright]

I loved the blurry early AI art. How things don't work in a metaphorical physical space like in a lot of AI art still today. They looked like the hallucinations I expected. Until then I haven't seen any art that looked like that. It's a pity that as time goes by, they get rid of all the cool quitks in favor of imitating human art.

[@andrewprahst2529]

It's probably a matter of showcasing new capabilities. But that doesn't stop anybody from using smaller models. What art is made is up to the artist and doesn't have to be dictated by whatever the latest trend is.

[@TrafficPartyHatTest]

That is my exact same stance on it, I loved the surrealist early AI stuff

[@mrosskne]

so make your own.

[@Hotshot2k4]

The unfortunate reality is that the debate of whether AI art _is_ or _is not_ art, for very many people, seems to have little to do with the question at hand. Instead, it's a proxy war over whether digital artists deserve to flourish under capitalism, or whether every opportunity should be taken to concentrate wealth into the hands of the controllers of capital, and those who invest in technologies created by the controllers of capital. The artists, understandably, are worried about the fruits of their labor being devalued and, if they rely on their art to put food on the table, whether there is any path for them in the future that makes financial sense. It is in their interest, and the interests of their supporters and general people-centric thinkers and leaders to argue that the whole of AI art is completely meritless, terrible, and attack it from any and all plausible angles, as well as at times implausible ones. The capitalists and investor class are excited at the prospect of new revenue streams and cost-cutting opportunities, and a further concentration of wealth. They hope to capture as much of the money that went to feed artists and their families as possible in order to enrich themselves, and some even take a seemingly sadistic pleasure in the suffering of marginalized artists that made art which is not to their preference or transgresses on their ideological beliefs. They will extoll the allegedly limitless virtues and possibilities of AI art, and repeatedly announce the death of the artist as a viable profession in order to hasten the transition, encourage more investment, and enrich themselves. The truth is somewhere between those extremes, but both sides broadly take a "if you're not with us, you're against us" attitude, and will argue that AI art is either The Antichris or The Second Coming, with no room for compromise in-between. It has become, in effect, a political issue first and foremost. In the long run we'll gravitate towards the truth, and it will be videos like this one which will help us get there, so I appreciate your analysis and thoughts on the topic. I am both excited for and concerned about the future of what AI will be, but it's very difficult to have a nuanced take on the potential of AI art during this raging social media storm.

[@Miraihi]

Probably the best take I've ever seen.

[@void_pepsi6405]

except that most online digital artists are hobbyists and for them it isnt a matter of labour/finance. your average deviantart or tumblr illustrator isn't worried about losing out on money they already don't make, yet they'll almost unanimously reject AI. it goes deeper than that

[@sunstone1957]

Yeah I literally just want to eat food and not have my own work used to put me out of a place to live. I would care less about this if it wasn't trained using the very same nonconsenting artists that it is meant to replace. The fact that someone can type in an artist's name to mass-produce work in their style with no compensation or even credit is disgusting. At least before people who stole our shit could claim they were giving us "exposure." Now the algorithm just leaves a bastardized squiggle where our signature once lived.

[@JDG-hq8gy]

The problem isn’t that automation is reducing demand for artists labour, but that the artists labor is being used to make the AI without the artists receiving compensation, which would normally have a neutral effect on artists, but since the artists are also suffering decreased demand it means that AI causes the production of real art to be under incentivised

[@JDG-hq8gy]

I disagree with the characterisation that this is all for the benefit of capital, people benefit from having the option to cheaply and quickly produce art

[@oscarguzman3017]

I'll admit that the first handmade painting that makes you nauseous in the same way that AI art does would be intriguing

[@caesarortega5249]

Cesar Santos does paintings like that. Normally he does amazing photo realism but he started making chaotic pictures that look vauguely organic like a AI hallucination.

[@mrosskne]

you've never been made nauseous by an image

[@thelocalmaxtax]

I'm reminded about the Brian Eno quote about how "Whatever you now find weird, ugly, uncomfortable and nasty about a new medium will surely become its signature" and how the extremely uncanny nature of AI art is in many ways the most appealing aspect of it to me. I often think about the "This person doesn't exist" photos where sometimes things go very wrong when generating the image, and rather than someone putting their arm over their best friend, it's a horrid amalgamation of flesh, teeth, and hair.

[@ramppappia]

ink the artist on tumblr has done some ai inspired pieces and used "badly" generated images in some pieces for this reason. I love her stuff

[@user-jq1mg2mz7o]

i miss that era of strange and uncanny fractal AI images... but alas commercialisation comes for all.

[@andrewprahst2529]

​@@user-jq1mg2mz7o I'm sure you can still use less developed models if you want to

[@mrosskne]

​@@user-jq1mg2mz7o there's nothing stopping you making your own images

[@squidmonk3j]

This entirely skirts what I'd assume would be main concern - that AI art is essentially dead labour.

[@Reionder]

If anything, AI taught me that small variations in otherwise common shapes/symbols can be utterly unsettling and I think artists could definitely use that in their favor. A badly AI-generated human face can be creepier than a mask in a horror movie

[@billpod11]

When you talked about how the photography in modern life has changed the way paintings are done all I could think of is Zetterstrand's art. It's so modern in a way that is refreshing. That some of his paintings could not have been made 30 years ago since they are based on aesthetics that came from certain video games. Notably using the spectator cam in Counter Strike to see through the map.

[@hermanr00]

On the other hand, there's Alex Colville whose works eerily predict old computer renders but predate them by decades...

[@juanlopezvillegas2967]

Just hypothetically. Assuming that an image generated by an AI can categorically be considered art. Who would be the artist behind the work? I agree that AI can become a very interesting medium for artists to explore, if it is just that, a medium. However, I think most people concearned with arguing about whether generated images can, or cannot be art, presuposse that on such claims the artist behind the work, or who creates the work, would be the artificial intelligence itself; not the human giving the ai a prompt. For instance, when you state that cameras can also produce disparities in results because of factors external to the control and intention of the photographer, these factors are also not determined by the camera, they occur within the world which just happens to be within part of the plane the photographer chooses to capture. And yes, such things as the maximum aperture of a lense or the sensitivity of a sensor are also outside of the artists control and are entirely part of the camera itself, but that would also apply to things such as the corrosiveness of pigment in oil paintings or the hardness of the rock available to sculpt. The plane captured within a photograph always exists in some way phenomenically, whearas the aspects within a generated image are entirely artificial, in many ways they are mimetic, they don't exist out of thin air, but they cannot be traced back to phenomenically perceived reality. I don't think anyone would say that then those factors in chance would make the world the artist behind a photograph, probably because at least as we phenomenically perceive that the world is not a sentient system that would have the intent of creating a thought out frame for a photographer to capture. For some reason that distinction seems a lot more murky and undefined when it comes to artificial intelligence, it is not sentient but yet it doesn't choose at random, and those "intentions" are not always designed by a programmer, for instance the certain ways it understands physical forms that distort in often commical ways. I ask this in good faith, I'm trying to defend a point, just genuinelly curious. Is there a distinction in claiming that an ai is an artist behind a generated work and that the world is the artist behind a photograph? Is the intention of a human writing a prompt enough to make him an artist, or if not what then creates the work? If there is no creator, no artist, no real artistic intention behind the creator, then can these images really be art?

[@akindlyorc]

I agree with this. People working with generational AI images are more producers or "idea guys" than artists. They are not able to impart intentionality and convey meaning without a ton of work that dredges up the process out of capitalist motives and plagiarism. Maybe a couple people in the world could claim to do this and they are not the drivers of the movement. It is not enough to make an image, you have to imbue the process with meaning which AI tools necessarily obfuscate. As soon as I find out that an image was generated by AI it instantly sucks out the defining feature of art for me. It becomes a husk that at best looks nice but has no meaning.

[@jackkendall6420]

I think that searching for an artist behind every piece of art is a dead end.

[@__D10S__]

@@akindlyorc "this time it's different guys! this new medium is actually not art this time! sure people were wrong about photography, cinema, recorded music, but this time, my natural inclination towards attacking new things like a white blood cell is actually justified!" people in 100 years will look back at these sentiments and laugh. as we do our predecessors. human society is just an endless cycle of this process. disruption, negative response to new stimuli, eventually normalizing to it, and repeating. art, technology, etc. it's all the same story. history goes on. we are not at the end of history.

[@akindlyorc]

@@__D10S__ I never said this _can't_ be an artistic tool but the driving factors behind it right now are absolutely preventing it from that. If you control the whole stack and process then you can start making that claim. No company is paying for this and most professionals can't or don't do this. This tool is being used for profit and novelty, not evocative communication.

[@__D10S__]

@@akindlyorc it’s been like 2 years. Dust takes time to settle. In the grand scheme of things this medium is in its infancy, and there is no reason not to expect it to mature.

[@Linguinesticks]

Something I've noticed a lot is people saying AI generated images aren't art as a form of insult. The word art has been coupled to a deeply positive connotation. To the extent that calling something a work of art is a compliment, and the reverse is an insult. So it's also important to ask what the word art is doing in this conversation.

[@LimeyLassen]

I think it's fair to say that it makes people angry, and saying it isn't art is a way of expressing that anger. I think arguments about how to define art are boring. The question why people are so angry about this is much more interesting to me.

[@Michelle_Wellbeck]

Traditional notions of Art no longer translates in contemporary society. It's mostly got to do with alienation from the traditional institutions of art whose situation is so separated from the common worker, requiring considerable education and resources to be able appreciate. The other contact is commercial and consumer goods which are kitsch. The last trace of real art in the sense of purposeless appreciation of cultural production is found in internet memes

[@vylbird8014]

I think the confusion comes from 'art' having two separate purposes, and so two related meanings. There's art in the cultural sense - art as personal expression, often deeply emotional. This is art for art's sake: The creation and viewing of art as its own reward. But there is also art-the-product - the business of art, where artists work either for the purpose of selling their work or on behalf of another that intends to profit from it. These are very different things. For the art-for-arts-sake, I think there's a lot of elitism from artists there. They've worked for years to hone their skills, and are proud of their natural talent. All that becomes worthless if skill is replaced by an app, allowing anyone with a little talent and a week of practice to achieve the same result. For art as a business, the fear is just that of unemployment: Experienced and talented staff can be replaced by someone fresh out of school and a Creative Cloud subscription.

[@renatonunez-melgarp2066]

and some of people complaining about they just make the most generic shit ever made

[@SadeN_0]

@@LimeyLassen it's also not a hard question to answer

[@_lucasmaciel]

Jonas, I think this discussion is missing an important detail. It's the fact that these AI iniciatives are often private (CAPITAL). Aren't their algorithms supposed to be immersed in new "wild" and yet to be known manifestations of ideology? How can "subversive" be born of something like this?

[@lxram9139]

Camera technology is oftern private as well

[@anondelon]

Many models are open source with permissive licenses

[@jonasceikaCCK]

I agree with this. AI technology should be open source as well

[@_lucasmaciel]

⁠@@lxram9139 Maybe the perfect example of the concreteness of the medium are Pinhole cameras...

[@tinkz2009]

​@@jonasceikaCCKA video on open source would be a delight!

[@MikeStoneJapan]

I'd really like to see something about not just ai and labour but automation and labor in general and the fact that cheap human labor is still widely used along side all our technological advancement

[@siddhartacrowley8759]

AI was supposed to replace work. Instead it replaces artist😔

[@danielsan901998]

AI is not replacing artists, it's replacing artistic work, the problem is that in the current economic system art becomes another commodity to sell, produced for work instead of pleasure and personal fulfillment.

[@zyrkugilgamesh]

@@danielsan901998 So we artists are doing it wrong by wanting to capitalize on what we do to make a living?

[@theangel666100]

​@@zyrkugilgameshmaybe. But really its the system

[@theangel666100]

"Replace work" i.e replace workers. Fustratibg that some people are OH so happy to replace working class people and professions without a thought but when it comes to artists now its a problem.

[@siddhartacrowley8759]

@@theangel666100 Fair point.

[@JujuAdams]

Anyone else agree that using subscriber names to generate AI images is performance art using AI and therefore an example of the sort of multi-disciplinary AI art Jonas is talking about?

[@jackkendall6420]

Yes, I thought it was a very elegant unstated way to emphasise the points made in the video.

[@gclip9883]

That is actually a great example of how we could use AI to do something entirely new. You may have seen these "Harry Potter but it's Balenciaga" videos, which i find absolutely fascinating. The creator used the AI in this case to create something that was very difficult to achieve beforehand. So there is something to be said about AI "art" with a more elaborate concept.

[@mrosskne]

which disciplines?

[@mrosskne]

​@@gclip9883it's quite easy to draw people in suits actually

[@hurdygurdyguy1]

3:30 ...I'm reminded of commercial advertising art in magazines in the '50's and '60's (especially car ads), the subjects were rendered more grandiose and spectacular than a photo ever could...

[@chacochicken]

Not Tendies man, damn

[@jorb1903]

I'm very confused by this argument. I'm not arguing about whether AI or photography are art, but theyre absolutely different. I find it really odd that your defense of photography copying previously existing art just like AI is to use an extremely specific example of photo collages. Is intentional use of existing photographs to create a cumulative work of art the same as an algorithm searching through millions of existing images to contour the result to what it was asked to do?

[@BinaryDood]

There are dozens of meaningful decisions in collage, not in prompting, which is so basic that it itself can be automated (or just copied and pasted from somewhere else)

[@jackkendall6420]

@@BinaryDood There are lots of significant decisions in prompting too. You have to choose your base model, your sampler, your temperature, your CFG, your negative prompt. There are plenty of products which boil down this simplicity to just the prompt, sure, but if you work with raw Stable Diffusion there's plenty of complexity to get lost in. Each of these knobs and dials has a significant impact on the final image (I speak from experience, having fucked up choosing the right sampler many times).

[@BinaryDood]

@@jackkendall6420 i know Comfi Ui. More knobs and levers but still mostly not up to you. Regardless, most use cases won't be that. Whoever cares about finding a "process" of sorts will be drowned by the near infinite output from those who don't.

[@jorb1903]

@@BinaryDood Interesting. So you're thinking about AI art in the sense of modeling the same constructive techniques that go into an artists' creatie choices as opposed to a simple prompt? Definitely something I haven't considered. If I'm understanding correctly, I guess whether or not AI art is "authentic art" (whatever that means), depends on whether those techniques themselves are "authentic." Also note: I was very drunk when I made that comment so I don't think I worded it as well as I think. It also came across more combative than I intended lol

[@JohnDoe-jp4em]

@@BinaryDood Control of what the generator does is essential when you actually have something specific in mind that you want to create. This is like saying finding a "process" to camera work is a waste of time because just pointing the camera at something with auto-focus and auto-brightness is much faster.

[@JMoore-vo7ii]

Would you ever consider talking about Berger or Benjamin on art's mechanical reproduction? So much of what is discussed in this video sounds similar to their work. I believe Breton was mentioned in one of their essays

[@philiphammar]

i'm deffo gonna reread that benjamin essay again. what was the John Berger book/essay called now again? so long ago i read it

[@JMoore-vo7ii]

@@philiphammar ways of seeing

[@fanrosefabrose9457]

R.I.P to Tendies123, fly high and soar beyond old freind

[@lemonboiyoutube]

what about intent? ive seen a lot of ppl make the argument that AI art cant be art bc the AI can't intend to convey a message. idk how much I buy that, but AI doesn't have perspective, or a concept of light which does make AI art often objectively worse

[@EduardoGarcia-qf4kg]

I think the lack of intentionality from the part of the AI is not important at all. Likewise, one could argue the mechanism of a camera does not have intentionality when registering the light that hits it. The obvious counterpoint is that the human being in control of the camera has intent, but this also allows us to argue that the human being prompting and choosing settings has intent and can pick one of the several generations that better convey what they will. but apart from this, I am not certain on whether it is productive to attach art to intent, as doing so implies there is something at the final work, some objective and knowable thing that intent confers that no mechanical reproduction ever could. I do not believe in such thing, it strikes me as wishful thinking, a desire to believe that there must be some sort of magical essence of humanity that leaks into our works and that such thing is objectively measurable. Even if we consider only human beings, this is complete and total nonsense. We can not know the thoughts of someone other than ourselves, we can not know the intentions of someone else. When we judge someone's intent in an art piece, we are guessing, and our guesses can be very good and accurate, as human beings can only act and think in so many ways and all of us have pattern recognition and plenty of experience to base our guesses on. But we are not accessing some transcendental plane of pure communication through art. Our guesses are subjective, not objective. The artist's idea and execution of how to convey their intended meaning is also subjective, as very often happens, they are misunderstood or no one "gets it" as they did. We can tell an image is AI generated because, as of now, they always commit some errors and follow some odd patterns (like not knowing perspectives or illumination), but those are practical problems, who might very well be solved in the future. In such case an AI could very well one day make an image that fools the entire human race into believing a human made it, and everyone could say "yes, this is human art, we can see the person who made thought so and so or wanted to convey this and that", what remains of the intent position now? Of what worth is this concept if it is completely useless at guiding actual human beings into identifying art? We could now salvage it by claiming that everyone is wrong, the image is NOT art irrespective of what everyone says, but that implies that there must exist some objective and true knowledge of what is art that remains true even if no one believes it, how could we even acquire knowledge of such concept, and why should we care about it? This is a funny thought, as many things we cherish as art might not make the cut under "the true form of art" that lives in the world of ideas. But anyways, I am not a Platonist, I do not believe in the real and true forms of concepts and ideas that exists independently of human experience.

[@lemonboiyoutube]

@EduardoGarcia-qf4kg i didn't really think that hard while making the message, but yeah i agree with most of your points. a few passing notes: it's just that in photos, the user has far more control over the end product. this means far more meaning can be had in a photo than a AI generated peice, since AI just takes a buncha data and "predicts" a similar peice. I'm not arguing that art needs to convey a message to be art, but the fact remains that the vast majority of art is inextricably tied to its artist. art is most of the time a medium for emotion, which AI currently has no conception of. ai art pieces are like the most generic possible conception of something, devoid of soul (i dont nessecarily believe in soul, but its a useful rhetorical device and yk what i mean). finally, AI fundementally doesnt have a conception of stuff like perspective, which makes its works painful and oftentimes lifeless. i think at best ai art, like the video suggests, will be used as a tool in the future. but the worst case, which is whats happening now, is people thinking it has more ability than it does, when there are inherent limitations to ai as it currently exists. i dont even rly agree itll be used as a tool akin to photography, theres simply too little control over the inner workings of ai as it currently exists for such art to be conceived of. you need human touch, touch which can give ai stuff like perspective in drawings, to actually achieve a satisfactory result rivaling current mediums for artistic expression. this is just my midnight rant tho so take it all w a grain of salt.

[@sphinxvibes416]

I loved this video and I appreciate you taking on such a controversial subject. I think something I've noted is that the most pushback against AI art seems to come from online fandom spaces who work in digital media. They're the most affected because they're digital arts (AI art couldn't really emulate the product of sculpturing or classical painting), but also I think because of how much of the art is tied to previously-existing properties or ideas by virtue of being fandom. AI art gets much closer to human capacity when the art is composed purely of references rather than concepts or ideas, which is reflected in how much of what AI is used for is other celebrities covering old songs, or drawing cartoon characters doing things. Its not that reference can't be used or communicate something new and interesting, but a lot of fandom art is inherently representational or built on consensus that can be emulated by AI. That also reminded me of your drum machine video, and your simpsons video where you cover the idea of a symbol losing all its contextual meaning and becoming purely self-referential. I am not blaming these artists for being mad, many of them are competing now with an inferior but cheaper way of doing their work, and the niche for experimental art is not nearly big enough in demand for them to survive in a consumer culture that seeks out rebooted properties. I'd submit the same is true for painters. Yeah, it 'liberated' the form, but it also made it a very small occupation. It all leads back to how capitalism tarnishes and restricts creative expression by tying it to the consumer, and that's sadly missed in discussions that frame something in ideological "real art" talk.

[@robertomonroe6338]

Because of my enmity for this technology, whose 'how the sausage gets made' mechanisms I know well as a technologist, I doubted I'd find anything intetesting in this video. I'm glad I watched. You make a good argument for what would constitute 'good' 'AI Art' - a departure from plagarism and use of software tools, *by artists*, to create new forms

[@mrosskne]

If AI is plagiarism then so is every work of art ever made.

[@mitakiharashi4367]

So glad you're uploading regularly again. I've missed your concise and complex thoughts.

[@asteroidalassassin6949]

AI art just proves most people have bad/trite tastes.

[@ethan1142028]

I really liked this video, and I think 10-12 minutes is perfect for some topics and 30+ minutes is perfect for others. Don't be afraid of shorter or longer content, as long as you think you are giving each topic the time it deserves and neither cutting too much nuance nor padding for extra runtime!

[@fernandorevilla3518]

This is amazing! I will defenetly be exploring these things! Thank you so much for showing me the posibilities

[@PauLtus_B]

While I think the discussion of the artistic value of “A. I. art" is very interesting, I also believe aggressively trying to discredit it as valid art distracts from the actual problems it causes. There's nothing wrong with people writing a prompt and an image coming out, the problem is people trying to replace actual artists with an automated rip-off.

[@SweBeach2023]

If their art can be replaced with an AI, are they really ARTISTS or just factory workers doing a job any monkey could do?

[@PauLtus_B]

@@SweBeach2023 I feel there's two ways in which "A. I. art" is appealing. The innocent way is simply to quickly realise something you have on your mind. The problematic way is just a cheap way to create a product that you can get money for. Aside from the monetary issues, the problem is also that it misses that art does not only exist for the audience, but also for the artist as a way to express themselves, that output is just gone then.

[@Howie_vibeMaster]

this such a great video, you brought up lots of point I never construed then thinking about ai art, glad I watched this

[@MantasticHams]

I appreciate this take, i've been an artist far before AI and i've taken quite an interest in it (a mostly solitary interest as i worry about the impression it may give people). I worked with photocollage, 3d, graphic design, typography, rudimentary cartooning, and ive had experience painting and using most mediums. For me it provides an avenue like you are saying to new visual possibilities. I'm generally starting with either a photo collage, a rudimentary digital painting, or a txt2image, but often some sort of combination of all those, and then i use many many iterations of inpainting, upscaling, collaging in new elements, recoloring, and using many different Controlnets to hone in on the details im looking for and create an extremely high resolution image that i feel i've invested a lot of personal thought and effort into, from concept to execution. Usually these images are in a variety of surrealistic modes, i like to (for one example) anthropomorphize objects in different ways, i like to go for sort of practical film creature styles like cronenberg films, but then accompany details from other incongruous elements, like anime facial features or hairstyles, childrens shows, claymation, classic surreal painters, pop art. I also like to take these and then heavily process and edit them, like editing them onto a film poster, or downscaling and dithering with a limited pallete to make it look like game assets. I also think there are new contributions to "glitch" aesthetics that help express the condition of the way culture feels right now. My feeling, to touch on the more sensitive parts of the subject, is that we need to worry about artists, but we also need to both consider and promote the fact that many, maybe most people are interested most chiefly in artists, not art, and in all those cases art made by people, to convey particular meaning/emotion/experience, will continue to be the most valuable commodity in the art world. What we need to hone in on is the scammers that will go after the artists, not the technology, which, when open source and not profiting the rich, i think is possibly quite beneficial to the future of humanity, or possibly just another relatively small mostly neutral development in history. We also need to go after the corrupt way that many non-profit groups are operating with for-profit branches, etc.

[@chriss780]

Tendies123... gone but not forgotten RIP

[@Sonji_S]

Art is a form of expression. It is essentially a language that works based on symbolic, often emotion based communication. This requires some level of intentionality. AI could be a tool of genuine expression, like how the brushes on photoshop create a digital randomness that can be used intelligently. But AI art as it stands now is too all encompasing and too passive to allow the artist any meaningful input

[@odb1612]

yeah i mean i could come up with a concept of an image, which has an intended meaning and then enter a very detailed prompt. in this case, the human made concept would make it count as art, but at the same time the result would look generic and lazy compared to any other medium. i think the only way to create good art with AI is to reflect on the medium itself or embed the results into a bigger artistic context that isn't just one generated image ag the end

[@zyrkugilgamesh]

@@odb1612 It's just a novelty tycoons are trying to capitalize due to FOMO. The real bad thing is how this unfinished tool is trying to be used for anything and everything, just because is the new toy and "full of new opportunities".

[@b_delta9725]

if you think about it, that's exactly how AI was progressing before its popularity spike. AI was used to allow some sort of randomness in specific areas of software, or at least something close to the human thought process, things generated from patterns we don't have access to. Now it became a form cultural phenomenon in on itself, many people create these images even as a sort of protest against people who hate it

[@iankrasnow5383]

Art doesn't have to be all about pure aesthetic and symbolic expression. That's kind of a modern bourgeois understanding of art. Before photography, you might commission an artist to paint a portrait of yourself or another person, because that's the only way you could record their likeness. A lot of art tried to be about maximum accuracy and not just aesthetics, because that was the only way to record an event or person's image. Nowadays, commissioning a painting of yourself or a loved one would be seen as an unnecessary luxury, something impractical that you'd keep just to show off. And so on. Humans have always done art, and most of the time it was to fit in with some practical purpose, not just so you could have something pretty to decorate your walls with.

[@b_delta9725]

​@@iankrasnow5383 but complex art of other things have always existed, you say "most of the time" it was done for practical purposes, but to how many things can you apply to? think about it, could the creation of handmade paper could be considered an art? if you look at videos that show how they do it, you could say yes, and the way machines create our paper nowadays isn't... but as a programmer myself i see the practice of design and implementation of exact structures as art, same as mathematicians cause it's art motivated by beauty, maybe the people who designed the paper machines saw their creation as artistic expression there's poeple who would rather do something else than draw people, but do it for the money, maybe there's people who don't enjoy the process but like the result, and they'd love cameras, there's all sorts of people and many just loved doing it, were obsessed and innovated in the art, it's all about how you treat it and that's why I speak about art in terms of meaning, there isn't much in the art of someone who did it only for cash, that wasn't the intention anyway, there is more intention if you put your heart and soul on it, and don't care if it's gonna be popular or even know, don't care if there's any purpose or practicality to it

[@joshuam8406]

Love these short videos as well

[@Gutsquasher]

This video was really well done and summed a lot of ideas I've been having about AI art and it's place as a new creative tool. The camera analogy is just perfect and really approachable. Defining the ends of the spectrum makes the middle ground easier to see too. Very rarely do either photographers or painters use a single tool. You have digital painters, photographers touching up their photos, it's hard to say what in these contexts is using the medium to its fullest extent. But it's very easy to say that bad art, whether it's made with a camera, a paintbrush, or AI, is barely utilizing its medium. And that is one thing that has irked me quite a bit. A lot of people are confusing a dislike for bad art with a dislike for AI. It's just a tool.

[@TwoForFlinchin1]

I think problem is that AI art requires "real art" to exist. Either you get funny hallucinations or you get a shit replacement for what already exists. It's idealized adoption in an industry would mean the replacement of artists and a functional end to the development of art.

[@double2helix]

I think you missed the mark here, to me what makes AI images and videos so worthless is the lack of an Artist. I have seen plenty of generated images that are very aesthetically pleasing and of which many interpretations could have been inferred, but there is no thought behind it merely an algorithm blending up preexisting artwork and presenting something that came of that.

[@willhart2188]

Every AI generated image still has an AI artist behind it too. The models have limits, but you can work around them to create new art by the process of experimentation, and by using AI as a tool to make your own ideas.

[@Americanbadashh]

@@willhart2188 No, you aren't an artist you are a thief and fraud. A genuine imposter

[@deadslashkill]

​@@willhart2188there are no ai artists, the program is producing an image, the prompter is more akin to somebody hiring an artist than an artist themself.

[@pygmalion8952]

@@willhart2188 so we should actually be grateful to those who thought about making art, or commission one rather than the person who made the actual piece? yeah caravaggio was merely a tool too for the commissioners. sure buddy. if caravaggio could have made his art 100000x faster would that make the commissioner an artist?

[@willhart2188]

@@pygmalion8952 I don't think AI art is a replacement for regular artist, just another way to make art. You should not use it for everything. I still commission regular artist and support them on patreon for example.

[@SSJKamui]

AI Art allows you to do tons of easy experiments and its pretty inviting for the surrealist method of automatic writing. I think I got the most fascinating results when I put passages from alchemical texts and compare the results. Once I got a picture of an anime anime girl with tons of red feathers in her hair, offering someone a cake which is literally burning

[@duncecaphero8334]

Always love your videos

[@quazarstars637]

AI is built to replicate what already exists, there is literally no avenue of originality as nothing it creates could ever be original, anyone who loves and appreciates art hates AI. There is not a single AI image that isn't a direct copy of a work made by a human that had countless hours of work and meaning put into it, if sculpting is art, AI is bad quality injection moulding, minus the effort. It is simply disposable, random, meaningless garbage.

[@PREEMGONK]

0:14 Holy shit louis It's Princess Jane

[@DeathAlchemist]

I am so happy you brought up someone programming and customizing the AI as a tool. It is something that would make AI more personal and allow an the prompter to take a more active role as an actual artist. At default, I see the AI process like someone telling someone at a restaurant what they want on their sandwich. You can give an order, but you didn't make the sandwich. From this, I discovered that what I like about artists is their own sensibilities and what they can produce within their limits compared to AI. The same can be said of photography tbh. It captures an image, but doesn't generate the image or the placement of objects in a shot, timing, and where you are taking the photo. Now these arguments against AI isn't me saying that its not art. If a urinal can be art, anything can given the context. Ultimately, whether AI can be art or is irrelevant to the problems that ppl have with it tbh.

[@varnull6120]

I'm glad to hear someone else make this point. There's something interesting to be made with AI image generation but it only comes by embracing the quirks of it. I played with a few self-hosted models and found that it makes some really interesting things when you give negative weights to positive prompts (and negative weights to negative prompts, making them positive). The machine kinda hyperfocuses on them in a way that creates a lot of distortions and brings out it's nature more - moreso than just flipping the prompts around. I discovered this by accident so would suggest to someone trying to do the same - write a normal prompt (girl, blond hair, drinking coffee) with the usual negative prompts (extra_limbs, bad_art, extra_fingers) then just add a negative weight at the end. If the line finishes with a weight it seems to apply that weight to the whole prompt

[@neozoen]

maybe the real AI art is the AI hallucinations we made along the way :D

[@user-jq1mg2mz7o]

i love the comparison of how painting evolved and expanded on what made it unique as opposed to being stuck with the very functional idea of "painting as capturing an existing image". and how current AI art sucks so much because it's constantly trying to be a fifth-rate version of something else, instead of things that are unique to it (for which there was a kind of brief moment early on with uncanny images formed out of specific 'mistakes' in weighting and image formation. but that's just one example!) (also love the callback to the drum machines!)

[@paulstaker8861]

Good AI art looks like normal art. Which is arriving at warp speed.

[@MaakaSakuranbo]

Per the argument of the video that's bad AI art, as it just replicated what already existed instead of using the new technology for something unique.

[@miguelignaciocarrere4486]

I respect Jonas so much that ill give him credit for making a good and solid argument about the use of AI art. However i must confess i'm quite worried about this. We hace a technological process that can imitate and reproduce (it's in it's infancy) almost all other mediums of expression that we use as humans. Even if the example of the history of the introduction of photography is interesting i can't stop thinking this is beyond a scale we can actually manage. Considering how many artists have been ripped of by now, and how the push to use this technology comes only from the drive of saving a few bucks... We will have a societal problem in some degree. It's not now, not even in a few years. But i worry about a society that leaves artists back or gives them a single purpose, to control and edit the outcomes of software...

[@Acsion42]

It was hard to stomach at first, but the photography comparison can go a lot deeper. When people were first introduced to them, they were hailed as the death of art, which sparked an overreaction in the opposite direction decrying them as not in fact art. Over time our understanding of photography has become much more nuanced, and now we recognize that it is a medium that can produce fine art, but we also recognize that not every single photograph is necessarily art. The same has been known for both writing and painting, as these media were invented and explored long ago. I wonder if people upon the advent of writing likened it to the death of painting? Or when ancient humans long ago first began painting on walls, did somebody chime in and bemoan that people couldn't appreciate just seeing things outside anymore? Maybe we can look at AI media in a similar way, and say that this 'commercial content' is not art, the same way a stop sign or an x-ray aren't art. These are merely tools, a means to an end, not the expression of a creative mind.

[@user-jq1mg2mz7o]

I was going to say one possible difference then and now is the attention of commercial interests/capital on the new image tech of that era… but now that I think about it, it was probably on the same relative scale

[@AConnorDN38416]

Never been so entertained by the patreons listed at the end of the video

[@FlosBlog]

THis was one of the most valuable contributions to the discussion that I came across

[@stratospheric37]

My favorite example for all of this is Mr Chedda. It's a funny image of a funny mouse, and made it into a good tweet. When people found out it was AI made, some artists tried to recreate the image and draw Mr Chedda themselves. But these new pieces of handmade art just didn't capture the essence of the original image, what made it funny good and interesting. Did AI art in 2022 make something irreproducible?

[@SadeN_0]

/dev/urandom has been making something irreproducible for the past 40 years

[@stratospheric37]

@@SadeN_0 Well this is something people have cared about, unlike random numbers

[@dkz8394]

I will sound extremely pretentious rn but hear me out - AI art used to be better before it was cool back in the days of the shitpostbot era of facebook there used to be a lot of bot-run pages that produced AI generated images, such as PerlinFieldBot 4150 (my personal favorite), whose titles and composition would be random based on a set of variables that the bot would mash together to create something, it wasnt the same type of AI (DeepLearning) but a more "rudimentary" one some people say that limitation is what create art in some ways, we can take this principle to understand that, working with the limitations of the deep learning method to create things as such as "Name One Thing In This Photo" and the three examples you showed at the end of the video, the weirdness they present create a totally new "aesthetic", similarly to glitchy generated images (having its own art niche through corrupting an image to generate a glitchy one). those things creat art pushing the possibility of the criative process of digital image to its limits and innovate by being thigs that people can reproduce but hardly come up with in the same manner this lead back to your example of the drum machines, where its being used to be a cheap reproduction of something that already exist, to cut corners and so on

[@CeramicShot]

Good post - this is one reason I still use Midjourney; it's extremely easy to use older versions of the tool if you want to. Being able to loosen or tighten its AI version of interpretive imagination via the "stylize" settings is another way it can avoid calcification to a degree.

[@T33K3SS3LCH3N]

The main problem with "AI" as we have it is that it's all trained on human materials. It's completely built to mimick what artists have already done. It can do some basic synthesis, but generally nothing that humans couldn't have thought of or done themselves. So as it stands I can only see three ways to use it for good: 1. t's ability to create media at a rapid pace or immense scale. These things are not particularly interesting on their own, but could be used in interesting ways. I can mostly see this in the context of video games/interactive media, where it can generate new outputs based on user choices. Designing a smart way in which how user actions influence the AI generations could lead to some very interesting experiences. 2. Enable more people to build projects that were previously too big for them, which could freshen up some genres of art that are currently reserved for big budget enterprises. 3. Find interesting ways to "break" it. As for works like at 8:40, we have to consider that there have long been hand-made generation algorithms. The "painting with code"-type of Generative Art (just with nothing worth calling an "artificial intelligence" in the circuit) has been around for decades now, and typical 3D modelling software also has capabilities that allow for the easy design of things that would be very hard or outright impossible in conventional art mediums.

[@FreshApplePie]

This is the kind of video I was looking for for a long time, I responded to another commenter about how I make art and it's tied strongly to my income so it's become a huge question for what the value of my art really is, but a question I was unable to answer.

[@LuisAldamiz]

The difference is that AI is not a tool but a creator, an inhuman one. You can use AI art as replacement from human-made art but you can't really use AI art as an artist.

[@Skullivon]

You absolutely can. Maybe not with rigid services like Midjourney, but the ecosystem that has sprung up around Stable Diffusion is an incredibly powerful toolbox for an artist willing to learn it. And professional artists WILL have to learn it, or something like it, to compete.

[@LuisAldamiz]

@@Skullivon - I'm probably speaking from relative ignorance admittedly, you caught me at "Stable Diffusion". 😅

[@Skullivon]

@@LuisAldamiz No problem! SD is a collection of open source models. They're freely available and can be run on a modest graphics card. You can direct the AI toward specific regions of an image, guide it with sketches or depth maps, and even make custom finetunes of the model itself to teach it new concepts.

[@thotslayer9914]

@@Skullivon so its a tool nothing more nothing less

[@danielsan901998]

Interesting the use of patreon supporters as prompt for AI generation, once you see AI models as the result of massive processing of internet data it becomes an interesting question how AI interpret less common concepts, since in a way it represent how the internet as a whole see that concept.

[@bugsephbunnin4576]

One of the things I enjoy the most of your videos is the part when you read the patreon's names. I don't know why I just felt it really satisfying and it's such an iconic part of your videos and you spelling names like gubgub-kolkol was really funny. Man I will miss Tendies123.

[@TheEvilmonkey25]

To me what true AI artists need to accept is that all good art takes effort. Good AI art can be made, but it has to be a display of the passion of it's creator. Right now the biggest hurdle AI art is facing is that it's marketed as easy and effortless and that just makes it inherently less interesting.

[@jackied962]

Sure but isn't much of Modern art seen as not taking all that much effort?

[@tenhayz1889]

The only way I see that AI could maybe generate actual art, is if the "artist" had full control over the learning process the AI goes through. Otherwise, it is just a black box trained to identify and replicate unknown patterns. It is like if a painter tried to paint, but was influenced by random things and couldnt choose his inspirations.

[@akindlyorc]

Without this control you're more idea guy than artist. We don't consider producers to be artists. Agreed 💯

[@SyoDraws]

The thing is that while something like "choosing your inspiration" and even "learning" is relatively intuitive for us, for the AI it is very complicated. Trying to show that to a human in an understandable way may be difficult But I agree: a lot of AI tools feel limiting compared to just making the thing yourself

[@elysios7521]

@@akindlyorc Yeah, its like you are a client and you ask the artist(AI) to create some art for you. You could take the title of art director if you are so inclined but you are delusional if you think you are the artist

[@Megaritz]

I'm inclined to think an AI-image is *more* an artwork insofar as the artist puts more of themselves (in some important sense) into the image, and *less* an artwork insofar as the artist puts less of themselves into the image. This can mean a number of things, but at least part of it would be exercising control over the prompt and variables in a thoughtful way and engaging in a lot of deliberation in how the final image turns out. One problem, I think, is that we often can't tell from an image whether the artist put "a lot of" themselves into it, or if they basically turned some dials in a low-effort or low-skill or low-thought way. This sense of the actual or likely "lack of the artist" in the image seems to me why a lot of AI imagery seems soulless. But to say the artist needs full control over the learning process seems an overcorrection. Painters don't have anywhere near full control over their own inspirations. They're influenced by many unconscious or poorly-remembered forces. Similarly, photographers have limited control over what ends up in the frame, despite controlling the frame itself (and sometimes some amount of what's in the frame, e.g. for staged photos). But painters and photographers produce art, despite lacking full control over inspiration or content. On the other hand, if the "artist" exercises *so* little control as to outsource the art to someone else, then it seems not to be a production of art by that person (but rather by the other person). On the other other hand, it seems sometimes (not always) plagiarism can be art. This is not to endorse plagiarism, but it means AI imagery is not shown to be not-art even if it is shown to be plagiarism. Natural formations, like mountains, are also not art (even though they can be aesthetically great). The necessary amount of artist's control for a thing to be properly art rather than non-art (and properly *that* person's production of art, rather than someone else's) seems to be non-trivial but less than full control over the inspirations or the content.

[@elysios7521]

@@Megaritz I replied when you were writing the message probably and you didn't read it but I will still stand by the one that is using an AI being at its best and art director but usually just a client asking for some art to be created by the artist which is the AI. Its like communicating with an artist with text and asking him for revisions. Rarely more than that. The only cases that the AI is used as an actual tool and is not the artist is when its used akin to photobashing/to render color some stuff

[@gaybowoser]

This is not really disagreeing with the video, i get the point and i really appreciate you didnt discredit more transfortaive expressions like collage, youtube videos, amv, ytp, etc. Just wanted to write some tangential thoughts (dont feel obligated to read them). For me, the biggest reason artists feel threatened by AI is what you said, AI is replacing artist to the point a lot of people cant even tell whats AI or not. It is really sad because it takes a long time to learn and we want to survive, but mostly im scared that people wont even care as long as the end product looks good to them. It feels existencially bad becuase i dont know if there will be more and more things that ai can create as easilly in the future (youtube videos, blockbuster movies, AA videogames, entire job fields), and less and less things for us to do, and im scared people wont even care (or can even tell) if its made by a human or not. I think no tecnological innovation or new medium is necesarilly unequivocally a net good, there are still a lot of loss, more bitersweet. I do get the framing of the video, photography as a "relief" from being shackled to representation or photorealism. But is it really a relief if you are being forced to adapt to survive? You could draw all of those more impresionistic things becase you wanted and not because you had to before. Probably a lot of portrait artist who actually loved their craft felt screwed that their art was now not a means to the end of anything anymore, it must have sucked to be one of them. I just hope that in the future people will still want to make things and appreciate us now making the things that we do.

[@CodyKendall1]

People wouldn't be using AI art if copyright wasn't overly restrictive.

[@gaybowoser]

@@CodyKendall1 what u mean?

[@DEGriffSoc]

Cool video. I am interested to see what work painters might produce inspired by these computer generated things. I remain unconvinced about the idea "AI" images themselves can be art. I think the programme itself could be art and all of the outputs a part of that single piece. But I am inclined to reject the images themselves as art because I am unconvinced there is an artist in the sense of an identifiable person whose desire and feeling and intent the image is a product of. The images are representative of no specific person's desire or feeling. Instead, what appears is an amalgam of all the input work, which produces a blur of deferred desire, the desire of the artists whose work was put into the system. Which work, which artists is filtered by the prompts but it isn't produced by the prompt. The feeling existed (long) before and was embodied in a work that is being obscured by the programme. Or, to put it another way, it isn't art for the same reason Adorno and Horkheimer argued Hollywood films are (usually) not art. Those films are not cultivated for art (to embody a desire, a feeling, a craft or an aesthetic impact), they're cultivated for profit, for maximum consumption, maximum palatability. "AI" generated images are generated to be maximally 'acceptable' because the goal is to get more users before any other goal. Though as mentioned, that does leave open the idea that the programming of these models could become an art. What work is to be put in? How is to be processed? What kind of prompts can it recognise? I think those kinds of questions might provide the basis of computer generated art as properly art. But as long as these models just take as much input as they can to produce the most flexible models that can satisfy the consumption of the most users, it isn't art.

[@Americanbadashh]

AI don't produce art, they eat art and produce feces

[@user-jq1mg2mz7o]

interesting point on how the programme could be art but not the output per se. kind of like how video games can be art but no one would call any one person's playthrough moment by moment art. an artistic-ish experience maybe.

[@DEGriffSoc]

@@user-jq1mg2mz7o Videogames are a really good comparison!

[@freedom_mayor]

I appreciate this so much. As someone who knows a little bit about the history of art and photography, a lot of the AI discourse reminds me of cameras and painting. painting was declared a dead artform many times, at first from cameras, but the invention of photography actually led to such a diverse landscape of painting because the medium wasn't tied directly to depiction anymore. idk

[@bdgnc]

Thank you for this very simple and eloquent video which i believe is really setting the stage for how we should discuss the topic of AI art

[@dafff08]

even as someone who generated a lot of ai po**, i can see how ai has some controversy to it. its like buying cheap goods. you know its been created at the cost of the workers, but you are still drawn to it. its a quite complicated topic since it involves many many new aspects. ranging from future job security, to the ethicality of mishmashing other peoples work and how it can even create new job opportunities. altho ai generated content has still many telltale signs that it is ai generated (the vector graphics like visuals/style), i think this makes it unique in it self. seeing your leaders getting transformed in to gym bros and spouting with ai generated voices gym slang can be quite funny. that said, another question arises regarding deepfakes. as for now most people can tell the difference between reality and ai generation, however what will happen with our media coverage once ai videos are undistinguishable from a real video. when it comes to ai i feel like we need the brightest minds sitting together and set boundaries that will benefit us, the people and not corporations. as things are evolving we are moving towards mass unemployment. however instead of being like, "oh snap im gonna be jobless", we should shift the focus towards, "neat im getting my life back". but thats a topic for another day.

[@SadeN_0]

that'd be real neat, but we all know it isn't going to happen. the worst aspects (for the worker, the cog, the underclass) always prevail.

[@websiteuser7926]

Thank you for finally making a video about AI art worth watching, the only other video about AI I liked (which had a section about art but was more about its culture) was the Jimmy McGee one and even he fell into the trap of denying the possibility of it "being art" (edit: it appears i had misread the video, reply below, also watch that video, it's great). I really do sympathize with it but I feel like at its core theres something reactionary about denying a new medium (even if most of what people are making is complete trash), I believe we will one day see a Stieglitz for AI and I cannot wait for that to happen

[@JimmyMcG33]

sorry for bringing this energy to a youtube comment, not trying to be rude i just didn't expect to see my name here. I just think that the 'culture at large' has failed to use AI in an artistic way, so I sidestepped the question and said that AI was not inherently creative nor a very good tool in its current forms. I imagine that's obvious for you and I--tools are not autonomous, they can't create anything apart from a labour input--but I was trying to counter the messaging from OpenAI and the like about AI sapience. There's a guy on KZbin--can't find him ATM--who explains machine learning concepts using AI-generated visuals and by taking on the perspective of the network in first-person as though it were a sentient being. His videos integrate the inherent horror and uncanniness of current-gen AI visuals in a truly fantastic and original way, by taking the sentience talking point to its conclusions he frames the AI's experience as something like psychic driving. It's visionary work. Alan Resnick also made some interesting stuff in the very early days of AI-generated visuals, but they probably aged poorly in the deluge that followed. In the video I even pointed out that Deadeye Deepfake Simulacrum was a good use of AI's inherent uncanniness (though I'm still not sure if the character portraits in that game are actually AI-generated lol). The possibilities are there, but reality hasn't really caught up. If artists had started independently training neural nets on their own work, the status of "AI art" wouldn't even be a question, in fact, smaller networks exhibit a lot of the dreaminess that people like in the AI aesthetic. What bothers people is AI's immediate deployment as a replacement rather than a new form--my local mall has AI-generated fruit stands pasted over unoccupied storefronts and it's like walking past a blurry nightmare of imaginary, rotting produce. Imo the reaction against AI art (and it certainly is reactionary, as you say!) comes from this gut feeling that the technology will be used to deprive us of good things rather than make more good things. For me, the most promising thing with AI-generated visuals (or whatever) is not the melty aesthetic but the possibility of prompting our collective Id in this hyperconnected form. I spent hours trying to get Midjourney to produce something worthwhile, but it's a poor tool for the job.

[@websiteuser7926]

@@JimmyMcG33 I actually really appreciate the reply! I hope my mention didn't come off as too dismissive or judgy because I really do agree with the central point of your video (most of the art stuff feels like a distraction to me in the midst of all the other issues it brings up) and I gained a lot of respect seeing you question yourself in the notes on your site when faced with the creativity aspects, I had actually planned to maybe email you about this at some point asking if you could elaborate on it. This fills in for what I assumed you might have decided to elide based on what really matters to the video. Thank you for explaining! Looking forward to the diablo 4 video btw :)

[@websiteuser7926]

Also, if you ever find that guys work explaining NNs I would love to see it, sounds very good. Alan Resnick was quite good at the time but I agree it probably hasn't aged well. Have you seen Conner Omalleys new work that incorporates AI? Its very subtle, more of a visual effect than something wholly "AI art" but the visuals are nightmarish and make it feel like a bad trip, great stuff.

[@atyboony6941]

2:33 little correction here, photo montage is when you edit the negative/the image, photo collage (colle in french meaning glue) is when you cut and stick images together

[@Starkus23]

Thank you so much. This video really hits the nail on the head. Coincidentally, I also just took a seminar on the Philosophy of Photography this past semester for my Philosophy BA, so many of the writers you mentioned were already familiar to me, which was super cool to see. A lot of artists are (maybe rightfully) worried about their jobs with AI art on the horizon. However, I think, another aspect worth mentioning in this whole "AI art debate", is the question of why we as human beings even create art in the first place. Do we a) create art in order for it to be consumed, in order to satisfy a certain demand of society, like we do with a lot of other commodities? Or do we b) create art because the creation process itself and the expression of ourselves in an aesthetic way is simply something that is an inherent drive of our human existence? a) would be the capitalist argument, which sees art as just another commodity. I personally however would rather assume that b) is the case. And if the creation of art, the production of art, to put it in economic terms, is already an end in itself, then there is really nothing to be worried about because that is something that AI can never take from you. Except of course your job, but that is of course a problem entirely created by the capitalist organisation of the economy and not by any means a problem about the ability of AI to create art now.

[@generalfishcake]

I work as a game artist. Here's a story to demonstrate how AI art is plagiarism by definition. My boss asked a colleague of mine to send him a folder of all the digital art he had created for the company in the past 9 years. The boss comes back several days later and shows off a Stable Diffusion model named after himself + the artist's initials. The model spits out hundreds of images similar to the artist's style. It was obvious from the outputs that the boss had also used the artist's personal work posted on ArtStation. On a deeper level, the base model is trained on millions of other paintings by world artists. So, the question is, what was the "author's" (my ex-boss') input in all this? Is he an artist, even if SD happens to spit out a very moving and interesting picture once in 200 outputs? How is this art, and not automatized plagiarism?

[@generalfishcake]

PS. If you're my interested, I can tell you the name of the company, since I quit last year and the NDA no longer applies. But it doesn't matter for the overall point.

[@priapulida]

the important bit here is "art he had created for the company".

[@elysios7521]

​@@priapulida It shouldnt be. Training an AI should have a different clause than copyright because it makes that artist's work and career obselete.

[@priapulida]

@@elysios7521 you can't regulate inspiration. Whatever the artist was inspired/influenced by before could be captured in some other way with different models, loras, prompts

[@elysios7521]

@@priapulida You are not regulating inspiration, you are regulating training data. This comparison of human inspiration with AI training is a bad faith one at its core and it was born by AI bros trying to anthropomorphize the AI

[@_bradleystrider]

i think there are genuinely artistic efforts being done with AI, usually in a surrealist vein. alan resnick of adult swim fame is doing some really interesting things with it, and other lesser known artists like petr válek as well

[@ClockTowerTitan]

I’m a professional artist and I had the exact same thought as Turner🌅 “Ai is the end of art. I’m glad I had my turn”. But this video actually made me a bit more comfortable with the idea of AI generated art..

[@not.spir0s]

i've had this idea for awhile, of using ai to get framing of a collage right, and then collaging together various things in photoshop and then painting the result to canvas. seems like exactly what you lay out at the end of your video.

[@jabberw0k812]

I've always been open to the opinion that photography is not art the way painting is. But even if it is, a photographer selects the image from what they can see. They look at a scene and decide what is worth preserving, or create a scene to photograph, and exert significant control over that scene in the process. Even someone using clip art has to carefully arrange it. Using an AI prompt is more like commissioning a photographer and giving them subject matter, then choosing which photographs you like. It's curating. It's the rough equivalent of claiming to be a scientist because you hired a team of scientists to work for you. Which is also something that happens. It *could* be a tool for a genuine artist to use in certain circumstances, but when the process is a black box, where's the line between creation and plagiarism?

[@Blaxpoon]

As someone who generated images, I can tell you you select a lot in the process. The whole point of ai is that you can create a lot of pictures very fast, so the main job of the human is less to come up with a good prompt than it is to select among the hundred of images it generates

[@pygmalion8952]

@@Blaxpoon "i select a lot when i commissioned an artist to make something that i want" and "i select a lot when i type a prompt to a black box" is the same thing. you are still a commissioner. you even pay money for it too. but way less per image. it is a market substitute created by stealing artists' own work.

[@htth3152]

From what I know some of the bigger AI users do kinda involve themselves in the process past merely choosing the result. Not as much as photographers, but they can for example jump between models on different stages for some intended effects, choose what part of the image to iterate on, collage together different generated images, overpaint them, etc. If this is also curating, wouldn't that make film directors, or any creative directors mere curators too? I think there's nothing _inherently_ un-artful about using generative AI as a tool. The reason 99% of the results is schlock, I'd say, is because the incredible ease of the process attracts a lot of "enterpreneurs" who simply hope to make a quick buck on this opportunity and have no understanding of what art (even normal art) is. They drown out those who actually use it artistically or cause them to abandon these tools because of the stigma. Unfortunately even the "bigger" AI users I started with are mostly these "enterpreneurs". Whatever artistic potential there is in this medium, it still has to wait for someone to truly utilize it.

[@fabiodeoliveiraribeiro1602]

Image generating AIs transform 000101010101, 10001111111001101 and 111000101110110 into 101010101000111. The result we see however will be very different, but beneath the surface of the image of 101010101000111 it will still be just 1010101010 00111. We can still distinguish pre-existing images from the digitization of the world. But in a few decades this may become more difficult. Everything will just be an ocean made up of 0s and 1s. In a world drowning in digitally generated, manipulated, regenerated and mutilated images countless times, will there be any space for art history, the study of aesthetics and artistic innovation?

[@BinaryDood]

No. All anchors of culture will be lifted.

[@MaakaSakuranbo]

Have you considered that this is artistic innovation and you just dislike it?

I would just add two things: the process of replication is actually important for improving the tool - its as if someone was learning to paint. Most great artist have to go through some derivative phase. Every medium that brings some new tools unlocking techniques and results that were previously very hard to achieve always produce much more bad art, or maybe just "noise". For each good photo there are millions of bad images that have no artistic value, they just sit on the disk taking up space. Another note I would like to add is that I dont think we should limit the scope of good AI art only for people that use coding and dig deep into the system itself (I am one of them). I think there is an arguemnt to be made for pure usage of prompt, that just generate images from words without any technical tinkering. I think we should be able to extract interesting concepts, ideas and images with proper use of creative thought.

[@tdjhue]

The ai art gag at the end was amazing loved it

[@BlanketSoreness]

I always looked forward to the tendies123 call

[@nathandrake5544]

I was looking forward to the tendies123 AI art :(

[@nohbuddy1]

Photography didn't put painters out of business obviously but AI will wreck things like animation studios, TV shows, and games. No studio will employ a person to be abstract when AI can do the same thing

[@jonasceikaCCK]

I think that's an overestimation of how much AI is capable of, to be honest! So far, predictions of how many creative jobs will be replaced by AI and automation have been greatly exaggerated. Media companies with a reputation to consider have so far at most used AI for highly limited purposes, and even then it often requires human supervision and adjustment. Anything beyond that immediately leads to visibly substandard results and tons of backlash. The exception is media that doesn't depend on any level of discernment on the part of the audience, like content for kids or crude clickbait. I agree that companies will always try to cut costs wherever possible but so far AI can only do that at the cost of a far worser product. Artists still have plenty to do just as painters did post-photography.

[@lucatriunfo8023]

@@jonasceikaCCK An overstimation? Or an understimation on your part? AI art is fast developing and we already had evidence of companies using AI art to cut cost. Denying or putting it under the rug like you're doing helps no one.

[@sebastianharley7159]

@@jonasceikaCCK Maybe so for now but you should consider that for many AI advocates and programmers getting rid of human artists IS the goal, whether or not they can produce any meaningful or good art is beside the point. Many have explicitly stated that they want to turn art itself into just another capital asset without having to go to the bother of dealing with the workers who produce it.

[@Americanbadashh]

@@jonasceikaCCK Jobs are already being replaced numbnuts

[@jonasceikaCCK]

@@sebastianharley7159 And they should be resisted. Artists should organize themselves to any extent they can.

[@theskinegg9168]

I really wanna do a challenge where a bunch of people paint something to look AI generated and whoever’s is chosen as AI wins

[@log_e555]

you have singlehanded changed my long-held opinion on this, that was some good thinking.

[@SadeN_0]

Comment section full of "prompt engineers" thanking Jonas for his efforts in legitimizing their taking a blender (or maybe a trash compactor?) to actual works of art.

[@jackkendall6420]

Perhaps art deserves the blender?

[@shakie6074]

Important to remember the Searle's "Chinese Room Argument", Flusser's concept of the "black box of photography", and Benjamin's mechanical reproduction, when approaching the reality of "non-human imaging". For working artists, I can see nothing but potential in the advent of AI. Following Legacy Russell and Deleuze and Guattari it is up to us - artist or not - to both "glitch" the commodifying functionings of technocapitalism, while also adopting and utilizing it as a tool for new creative aesthetics / lines of flight / etc. Like Deleuze said: “There is no need to fear or hope, but only to look for new weapons.”

[@KoruGo]

Walter Benjamin's essay is one of my favorite texts on art of all time

[@MalditoSeasEstadoDelsrael]

Ai is art as a product, it can only reproduce what already exists mincing previously existing works together but without it's aura or something that's more than the sum of its parts, it lacks the human intent behind representations. Just take Borges for example, he was also a reader before a writer and his stuff is an amalgam of all the European literature, but written from outside of Europe, from an unique subject from a different context, and that's what ai will never have.

[@vylbird8014]

That's true of most art though. Very rarely does an artist do something really innovative.

[@odb1612]

@@vylbird8014a human made synthesis of previousoy recieved works will always be more than the sum of its parts, since all his experience, personality, emotions, thoughts etc are part of the decision how said synthesis will turn out. an AI will just always make an onthologically inferior derivate of previous images without being able to add some reflexion on the human experience to it

[@pygmalion8952]

@@odb1612 well said. it does not have any experiences. why anyone should care what comes out of it?

[@amirhossein2820]

I loved your take on AI arts. Your videos are always very informative and interesting. Would it possible for you to make a video on AI fiction writing and the possibility its outsourcing writers? I would love to hear your opinion on it since it seems to be looming over the future of most writers.

[@louisparry-mills9132]

Fantastic video.

[@estebansanchezoeconomo6827]

Your take is interesting (I greatly enjoy your videos), but I'm still skeptikal. I'm not sure that a comparison with photography is completely fair: you cite some authors from the 19th Century, who reasonably doubted the artistic capabilities of a new medium. Yet it became very quickly evident that the process of taking photographies is inherently subjective, and we must agree on the fact that a photograph is an original human product, even when it's not art. In the case of AI, without tweeking and intervining in the code, prompting is a much more random and disconnected input than producing a photograph, and the fact that the result is a mechanical and statistical combination of previously existing human productions just adds to it. The fact that AI art can be imagined or expected to come, does not mean that what exists at this point can be considered as art, and at least not connected to the realm of creativity. 19th Century painters did not have to consider that photography was actively using their productions, and phtotogoraphy became an art when it escaped from the codes of academic painting in the 1870s. As of today, there are already artists using AI in pretty creative manners, but to my knowledge, always with very limited technical aptitudes. Maybe a comparison with audio samples could be interesting: in the future it will be a matter of who can better prompt.

[@ZtlaMusic]

Good video, enjoyed it! To me comparing the lack of control present when prompting an LLM to the lack of control when taking a photo felt a bit off, though it's difficult to put to words. If one wants to take a photo of a subject, if they fail, the result is a failed photo of said subject. But if one prompts an AI (of the kind currently out at least) to make an image of a specific subject, the failure is basically guaranteed - but the failure likely is not a poor execution of what the prompt engineer had in mind but rather the wrong image. Creating art with an AI inherently involves accepting the fact that if you have a clear vision of the kind of artwork you want to create, you likely won't be getting that, but rather hopefully something quite similar. Unlike with photography, the lack of control with AI art is not about not being able to fully control the subject and circumstances but rather about not being able to fully control the actual vision.

[@keyworksurfer]

incorrect. prompting LLMs is for casual newbies, and no serious AI artist does it. using modern tools like stablediffusion and openpose, one can exert a great deal of creative control over the outputs beyond mere prompting: shot composition, posing of characters, placement of contours and folds, etc. learn more. speak less.

[@ZtlaMusic]

@@keyworksurfer If you think vague control over those factors equates control over the vision of the piece, then I guess we have different definitions for what vision means. As long as there is a gap between what you'd like to see and what gets delivered my point stands. As I said the best you can hope is kinda close, but if you've ever actually created art you know just how detail oriented it is. Agonizing over seemingly miniscule details is what makes art personal and if there comes a point where you'd like to make a change but cannot get the AI to make that specific change in that specific way no matter how well you know the tool, the gap exists. I think it's worth clarifying that I'm not trying to make the case for ai art not being art. I'm much more reserved about calling the people making it artists, though depending on the methodology I wouldn't necessarily be opposed to that either. I merely was pointing out that comparing the lack of control of inherent in photography to the lack of control when asking an AI to do something for you felt off.

[@sisyphusinboots]

the first and maybe only use of AI for serious artistic purposes that I was genuinely impressed by succeeds for basically this exact reason. It's the album video for Lebanese sound/video artist Yara Asmar's "Synth Waltzes and Accordion Laments", it thoughtfully combines real footage with AI for intentionally surrealistic dreamlike purposes in a way that feels more like drifting in and out of sleep than cutting to a weird CGI part, I guess I would compare it to how a movie like waking life used rotoscoping, albeit to a very different visual end. there's some weird puppet stuff too but somehow it just works, devoid of irony. for some reason she seems to have unlisted the full video and I don't want this comment to get axed by posting a URL but if you search her name and poke around her website it's still embedded.

[@gustavocortico1681]

My opinion is that good AI art will come from multiple models chained together in a way that resembles or straight up copies an artist process and/or the collective work of multiple artists and intermediaries.

[@guilhermeoutro6083]

Jonas, I think you missed the point comparing the critiques about painting x photography x AI art. The presence of mechanism or "automatism" doesn't make photography closer to AI art, since in photography the creative aspect is yet in the hands and eyes of the photographer: the positioning, choosing the perspective and the angle, using the vast array of technical apparatus etc. AI art, in the other hand, is much more like _ordering_ something to an external employee: you type instructions and then the machine presents to you a variety of results determined by an algorithm. I don't become an artist by asking someone or something to do this or that, but by actively doing it, even if this action is mediated by a technical apparatus (and very often it is). A photographer doesn't type instructions to a drone and wait for the results, they take the photo by themselves, using a camera. What concerns me about AI art is the devaluation of human work that comes with it: we shouldn't forget that this technology is not neutral, it's another way of equalizing things, of replacing and stealing people's creativity and then selling it as something devoided of uniqueness. In other words, it's another horrible development of capital and merchandise, the society of spectacle 3.0.

[@philiphammar]

yeah exactly. to compare ai generation to photography misses the point completely in my opinion, and i don't see why people keep on bringing it up honestly. it's these ai's coming up with new (in part stolen) IDEAS, concepts and motives that scares me personally.

[@jirivegner3711]

One of the potential use case for the AI, I am most interested in, in the "visual art sphere" is an ability to train it on concept art and use it to generate tons of coherent assets. If combined with few handcrafted unique key pieces, it could be used to procedurally create pretty decent looking fantastic words and either use them directly or as base/placeholder to improve upon. This seems like a potentially really interesting job on the boundary between art and technology. Kind of concept artist on steroids, whose job is not only to create/select/curate concept art, but also to train ai on it (plus whatever additional material is needed) and fine tune it. This specific AI would IMO both tool and a piece of art.

[@ZombieBobSponge]

I'm an artist. To me, seeing AI-generated images has only ever left a negative impression. They look cheap and lower the overall value of the medium. However, I agree with you. If we can find a niche for AI that only AI can do, similar to photography, that would be great. Interesting and thought-provoking video!

[@Tubeytime]

This is probably the best take on AI art I have heard so far, and I have said that exact sentence before!

[@sedarkaym3266]

i think the argument of is AI art or not is completely useless.any argument to say its not art can be met with an example of another artform having a similiar quirk. thats not the issue with AI images, its the fact that its both plaigirizing with no consent or recompense and its replacing many digital artists in the workforce. in an egalitarian society, the invention of AI images would be wonderful as it would alleviate the really tedious parts of the job but under capitalism its used only to furhter exploit the working class artists, reduce their demand, reduce their wages, reduce their rights and have them still work just as hard if not more because advancements in productivity under capitalism doesnt mean less work but rather the same amount of work but way more profits. as it is right now, AI only helps corporations and is an enemy of the working class artists.

[@BinaryDood]

Oddly enough, should you replace "is it art" to "is it creation", the argument looses so much of its abstract veneer that it is forced to consider actual in-world ramifications.

[@SadeN_0]

it seems to me that your piece intentionally disregards the major issue of how art, and who gets to do it, fits into our capitalist reality. the intellectual exercise presented would fit well in a post-scarcity utopia, but we are not there. what we do have, instead, is a technology that took a steaming shit on real people's real livelihoods, and entire communities are in crisis and hurting.