26 votes

An AI-generated artwork won first place at a state fair fine arts competition, and artists are pissed

61 comments

  1. [15]
    teaearlgraycold
    Link
    This seems like the same state that video game speedruns were in 10+ years ago. People generally considered runs with glitches to be "cheating". Eventually the debate was solved by splitting...

    This seems like the same state that video game speedruns were in 10+ years ago. People generally considered runs with glitches to be "cheating". Eventually the debate was solved by splitting leaderboards into "glitchless" and "anything goes" sections.

    The artwork in this article was itself in a "digital art" category. We've already decided as a society that digital artwork is sufficiently different from traditional art that it needs its own category. Art isn't judged just on its appearance, but on the effort required and means by which it's created. There's no reason we can't have a new category for "AI assisted" artwork. The artist will still bring their skills in composition, theme, style and subject matter to the table. AI artwork is more like photography. A photographer does not place pixels on a canvas one at a time. They find a subject and setting, camera angle, lens, etc. that combine to create a final work. It may take less time than creating a photo-realistic painting but no one would bother comparing photography to painting. It's not cheating because you're not even playing the same game.

    31 votes
    1. [14]
      MimicSquid
      Link Parent
      I like this take. It leaves space for respect for the person involved in the process, be it the person who spends 5 years to make an enormous sculpture out of hand-placed toothpicks or the one who...

      I like this take. It leaves space for respect for the person involved in the process, be it the person who spends 5 years to make an enormous sculpture out of hand-placed toothpicks or the one who takes a chainsaw to a log and gets a result they're happy with in 15 minutes. One took more investment of personal time and effort than the other, but they're both art.

      7 votes
      1. [13]
        Rudism
        Link Parent
        I can respect the fact that doing certain things used to take a lot more work and skill--editing movies before digital film; writing computer software using punch cards; heck even just writing in...

        I can respect the fact that doing certain things used to take a lot more work and skill--editing movies before digital film; writing computer software using punch cards; heck even just writing in general before typewriters, electricity, and our modern education systems was probably something of a novelty. But anyone doing any of those things without modern conveniences today would probably be seen as an eccentric, rather than some kind of auteur who is somehow more refined and "better" than the masses doing essentially the same job but faster because they're embracing the latest tools designed to help them do it. The painting that won that competition isn't like a log-and-chainsaw to toothpicks comparison--if you told me it was hand-painted over the course of years by a brilliant artist I wouldn't blink twice. It looks (to me, at least) really good and at-a-glance indistinguishable from something a human might have painted. I think in a hundred years we'll be saying stuff like "Timmy, did you know that paintings used to be made by humans by pushing sticks of colored goop around on paper?" the same way we talk about punch cards or kinetoscopes today.

        3 votes
        1. [7]
          teaearlgraycold
          Link Parent
          I don’t think so. AI generated art is not an objective upgrade as typing was to punching cards. People will want control over brush strokes rather than selecting whole images, or chunks of images....

          I don’t think so. AI generated art is not an objective upgrade as typing was to punching cards. People will want control over brush strokes rather than selecting whole images, or chunks of images.

          At least going from what is displayed in art museums (which is a curated view of history) it seems that painters collectively lost interest in photorealistic paintings after the invention of the camera. Why chase perfection when any old fool can get you perfection with the press of a button? Only you can use your human mind to give a subjective impression of the world.

          AI art will eliminate the need for utilitarian imagery. Stock photo websites will die. Many corporate artists will lose their jobs. But people pay extra for original paintings when cheaper prints are available. And picky artists will want to have the ultimate control over the final image. People will still be making traditional art and people will still buy it 100 years from now.

          8 votes
          1. [3]
            stu2b50
            Link Parent
            In that vein, X years from now, being an artist, at least a traditional one dictating every stroke, as a career will be thought of the same way being an NBA player would be. There will be a...

            AI art will eliminate the need for utilitarian imagery. Stock photo websites will die. Many corporate artists will lose their jobs

            In that vein, X years from now, being an artist, at least a traditional one dictating every stroke, as a career will be thought of the same way being an NBA player would be. There will be a handful of artist whose skill and story will be such that the weird but very real humans desires for intangibilities will allow them to be famous and their works to be highly valued.

            Outside of that maybe 100 sized group, artistry will be purely a hobbyist endeavor, because it will be impossible to monetize, the same way that the skill range of baskeball from YMCA to NCAA is also unmonetizable - you need to be good enough to be in a pro league to make money.

            It'll be a pretty big shift in the labor market for creative works.

            5 votes
            1. [2]
              papasquat
              Link Parent
              It's pretty sad. AI Automation has kicked off a desperate race to the bottom, just as automation has in most other industries. Master furniture makers are few and far between, because paying a 50x...

              It's pretty sad. AI Automation has kicked off a desperate race to the bottom, just as automation has in most other industries.
              Master furniture makers are few and far between, because paying a 50x markup on furniture doesn't make sense when you can go to Ikea and get CNC mass produced cardboard crap that looks passably decent enough. Local farmers don't make sense when you can get a slightly less fresh factory farmed chicken for half the price. It doesn't make sense to pay a copywriter to write decent copy for your product or article when you can get an algorithm that gets 95% of the way there but has strange quirks and the writing feels hollow and bland.
              I see art going the same way, where paying a graphic designer to make a logo for you or paying illustrators to create art for your story or product
              I get the market forces at play here, people want more stuff, and they want to pay less for it, and the loss in quality is negligible. It still makes me extremely sad though. It's just one more area where our lives are dominated by mass produced, automated, cheap crap.

              6 votes
              1. Akir
                Link Parent
                To be honest, I think the kind of art that will be replaced with AI is already pretty much the bottom of the barrel stuff, the kind of thing that you would get for 'free' with a stock photo...

                To be honest, I think the kind of art that will be replaced with AI is already pretty much the bottom of the barrel stuff, the kind of thing that you would get for 'free' with a stock photo subscription or on services like fiverr where people from much poorer countries accept relatively tiny fees for their work. As it is right now, "corporate art" is fairly reviled among the public, so one slight benefit to this loss is that there's going to be less of a chance you'll be hired to make something you hate.

                While AI art is a pretty nifty tool, we can't forget that it's still just a tool. It may be be better at "painting" than you are, but it'll never have ideas of it's own, and so there's plenty of artistic jobs that won't be taken over by them. I've been playing with Dall-E for a while and one thing that it absolutely can't understand is abstract ideas. If you tell it to make you a picture that is happy, it might be smart enough to use a lighter color pallet or to draw subjects who are smiling, but it doesn't actually know what it means for the picture to depict that emotion. And when you move to more complex ideas it may not even understand what you mean. By itself, it probably won't be able to come up with a clever and interesting logo design.

                1 vote
          2. [3]
            Rudism
            Link Parent
            I don't see any reason why, at some point in the future, AI assisted art wouldn't include the ability to control individual brush strokes in the generated images, and other forms of fine-tuning to...

            I don't see any reason why, at some point in the future, AI assisted art wouldn't include the ability to control individual brush strokes in the generated images, and other forms of fine-tuning to get the final output looking exactly how the artist wants. It's a crude tool right now but it's only going to get better and better now that the toothpaste is out of the bottle.

            1. [2]
              teaearlgraycold
              (edited )
              Link Parent
              You can already give it an Ms Paint sketch to guide the composition. But at the level where you want ultimate control there’s no point in AI assistance. If you already want to place every brush...

              You can already give it an Ms Paint sketch to guide the composition. But at the level where you want ultimate control there’s no point in AI assistance. If you already want to place every brush stroke what good is AI for?

              1 vote
              1. skybrian
                Link Parent
                Making a realistic preview from a sketch might be useful even if you plan to do it again yourself. And I also expect that some brush strokes are more important than others; maybe not all of the...

                Making a realistic preview from a sketch might be useful even if you plan to do it again yourself. And I also expect that some brush strokes are more important than others; maybe not all of the preview needs to be redone?

                It might be useful to compare with animation. Even for traditional cel animation, I think coloring is pretty well automated?

                1 vote
        2. [5]
          papasquat
          Link Parent
          There's one major limitation of AI art though. Its inherently derivative. Every piece of decent AI generated art looks like something some human did before, because that's what they're trained on....

          if you told me it was hand-painted over the course of years by a brilliant artist I wouldn't blink twice. It looks (to me, at least) really good and at-a-glance indistinguishable from something a human might have painted.

          There's one major limitation of AI art though. Its inherently derivative. Every piece of decent AI generated art looks like something some human did before, because that's what they're trained on. They'll never create anything truly original, at least not with the technology we're currently using. If "AI Art" gets lumped into the same category as human made art, the art world would very quickly coalesce on the new lowest common denominator, because why put forth hundreds of hours of effort to create something truly original if you could click a button and have a computer do it in thirty seconds?

          The thing that would be lost though, is true human creativity. If Picasso used AI, cubism wouldn't have happened. If Matisse used it, Fauvism wouldn't have existed. Computers are, and always have been extremely good at following instructions. They're recently gotten better at following very advanced instructions, resulting in the ability to ask a computer to draw a painting that looks like a specific artist and getting a passable result. They're still just following instructions though. The instructions now just happen to be a combination of their prompts, and their training material. There's zero actual creativity going on though. A computer will never have a beautiful thought and be inspired to paint the way it made him feel, they'll never suffer the depths of depression and express that on a canvas. They'll never even see a mountain stream that seems particularly beautiful to them and then try to capture it. They just do what they're told to do. A human being prompting a computer to do these things may have some sort of artistic value, but it's not at all the same thing as an artist creating from scratch a piece of art.

          AI Generated art taking off as a replacement for actual fine artists would just result in the at large art world become stuck in an eternal stasis with nothing really new ever being invented, which is pretty sad.

          3 votes
          1. [3]
            MimicSquid
            Link Parent
            There's one major limitation of human art, though: it's inherently derivative. Every piece of decent human generated art refers back to something some human did before, because that's what they're...

            There's one major limitation of human art, though: it's inherently derivative. Every piece of decent human generated art refers back to something some human did before, because that's what they're trained on. They'll never create anything truly original, at least not as long as they're trained using the art that already exists.

            Right? It's not like humans actually have truly original thoughts based on something other than their surroundings. Some art is a refinement of what came before, some is a refutation. But they both are responses to the existing corpus. And as long as AI art is based on human prompts or desires, those same forces will shape art.

            6 votes
            1. [2]
              papasquat
              Link Parent
              That's the thing though, AI isn't currently capable of making refutations of art that came before it. Its only capable of imitating it.

              Some art is a refinement of what came before, some is a refutation.

              That's the thing though, AI isn't currently capable of making refutations of art that came before it. Its only capable of imitating it.

              1. MimicSquid
                Link Parent
                Nah, you just turn down the setting that tries for faithfulness to the source material and it spits out something very different. Not that any individual one will be good to start, but sufficient...

                Nah, you just turn down the setting that tries for faithfulness to the source material and it spits out something very different. Not that any individual one will be good to start, but sufficient for the person choosing the prompts to latch onto and iterate on until it makes something great? Yeah, certainly.

                1 vote
          2. rich_27
            Link Parent
            I disagree. I think that is a quite a reductive view. In any given piece of art, the artist and the medium combine to create the piece, and the result is greater than the sum of it's parts. If you...

            I disagree. I think that is a quite a reductive view. In any given piece of art, the artist and the medium combine to create the piece, and the result is greater than the sum of it's parts. If you consider them in isolation, you aren't really taking a fair view; take a 3D printer for instance: to dismiss it as an XYZ plotter with no ability to create or make wonderful sculpture in isolation would not do it justice, just like writing off a car because it can't drive without someone at the wheel. In all three cases, the person is adding value, which makes it possible to do something the artist could not do on their own, be that creating a 3D plastic sculpture, traveling 60 miles in an hour, or turning words directly into imagery.

            Current AI generated art is a tool, and one capable of doing some amazing things in a ridiculously short amount of time. However, it is simply a tool at this stage and the person adding the creative element is the person guiding the brush, the artist who specifies the prompt in all its nuance and who selects the artwork from the generated set. Not only do they apply the paint to the brush (wording the prompt), they guide the strokes (choosing the subject matter, composition, and styling), and execute exactly the same discernment in deciding what is good and what is not worth displaying.

            I think the real crux of the matter isn't that any worth or value has been removed from the process with AI generated art, it's that the divide between the artist and the tool has shifted. All the same steps take place, but where previously the artist had to execute their ideas on to their medium, now the execution has shifted to the tool and the artist is merely the guide. If I sit holding the paintbrush, mixing the colour you tell me, and letting you move my hand as you choose, are you not still the artist? If you take one step further back and just tell me which bits to paint and how to move the brush, are you not still the artist? Were the masters of old that outlined their work before having their students fill in the details not still the artist, does the work not still bear their name?

            If we look at it from the perspective of a future AI system that is both the artist and tool - a program that generates art without prompting - that would require much, much more depth. That would require creativity and experience from which to derive art. I don't think it's fair to even consider current AI art generators in this frame, because they are simply not capable of that and to reduce current AI art to the product of algorithms and a combination of existing works completely removes the artist from the equation. Sure, current computers can't have beautiful thoughts or paint from how they feel, but arguably, something that can is no longer a computer. I don't believe we'll never see electromechanical life that can think or even feel, but if we do it won't be a computer, it will be something new.

            The thing is, any huge change is going to upset the status quo. I'm sure the horse breeders were devastated when the motor car grew to be the default mode of transport, but that doesn't mean it's a bad thing. In the immediate future, I'm sure we will see a flood of new art that is often hugely derivative and doesn't stand on its own as something truly new, but that will just be a reshaping of the landscape. When fine art is a dime a dozen, the things that really wow you, the real creative spark or the piece that speaks to you will become a diamond in the rough and that will become the new definition for what in art is fine. Yes, it might make things difficult for a lot of people whilst we reach a new normal, but that doesn't mean the new medium is in-and-of-itself a bad thing. It should also be noted, we are the sum of what we have seen and experienced. All fine art has it's roots in the things that came before it, however subtle that influence is. You could look at the first time someone used a palette knife to spread paint as a bastardisation, removing the brushwork and the subtlety from the equation, but it being different doesn't make it worse.

            I'd be interested to hear you expand on "A human being prompting a computer to do these things may have some sort of artistic value, but it's not at all the same thing as an artist creating from scratch a piece of art.". I really struggle to see your reasoning here, because - to me - all I see are similarities. Whether we see it in what we make or not, we draw our inspiration from what we have seen, and I'd argue the majority of art produced solely by human hands borrows extensively from works they've seen before. Who's to say an AI art generation program solely pulls from existing art, either? I could imagine art programs being trained on wider datasets than just traditional art. Even AI art generators with training sets including previously AI generated art would create something new. I think it's important to keep in mind that current art wouldn't be possible without the incalculable number of hours that have gone into designing and improving paintbrushes and other media, and, likewise, another factor to consider is the work and the creativity that goes into designing a program that uses AI to generate art. To me, once you step back and include the whole process - designing the generator, selecting the training data, monitoring and shaping the resulting program, concepting the piece, crafting the prompt, tuning and tweaking, and selecting the piece - there is a huge amount of creativity and soul that come together to create the piece. It might not be the same as traditional art, but, in my mind, it's just as worthy.

            3 votes
  2. [28]
    cfabbro
    (edited )
    Link
    No. No it doesn't. Because anyone trying to compare setting up a generative art program and learning to effectively write prompts for it (which takes a few hours at most) to the skills required to...

    “How interesting is it to see how all these people on Twitter who are against AI generated art are the first ones to throw the human under the bus by discrediting the human element! Does this seem hypocritical to you guys?”

    No. No it doesn't. Because anyone trying to compare setting up a generative art program and learning to effectively write prompts for it (which takes a few hours at most) to the skills required to physically produce similar quality artwork by hand, which can potentially take a lifetime of dedication, and a hundred plus hours of effort per piece being created, is delusional to the point I suspect they are nothing but a troll.

    If he had been the one to actually create MidJourney I might have had more sympathy to such an argument, since that actually would have taken a similar level of dedication, skill, time, and effort, albeit in programming and not fine art. But he didn't. He simply joined the beta, played with it for a while, and then used it to generate something to win this competition with.

    p.s. And fuck him for doing this too. Fine art is already hard enough to make a living from despite all the dedication it requires. And IMO wankers like this now trying to take away one of the few places where an artist can finally gain some recognition, and hopefully even win some prize money, is absolutely despicable.

    13 votes
    1. [4]
      post_below
      Link Parent
      I'm not sure that's fair. He labeled his submission "via Midjourney". I read it as an experiment. He might be a dick, but I don't think we have enough information to tell. I think maybe the fuck...

      And fuck him for doing this too.

      I'm not sure that's fair. He labeled his submission "via Midjourney". I read it as an experiment. He might be a dick, but I don't think we have enough information to tell.

      I think maybe the fuck you is better aimed at the technology itself, because the day it was released this kind of thing (and so much more) became inevitable.

      I relate to your sentiment, I've created time consuming art, and I have alot of respect for professional artists and the magic they make.

      But also, there's no putting this back in the box, so we're all going to have to make peace with the implications.

      The power of machine learning powered AI is going to grow exponentially for a while.

      13 votes
      1. [3]
        cfabbro
        (edited )
        Link Parent
        Oh, come on. He labeled it "via Midjourney" but do you honestly think that anyone in the judges panel at the Colorado state fair recognized that for what it was an admission to, or that it was a...

        Oh, come on. He labeled it "via Midjourney" but do you honestly think that anyone in the judges panel at the Colorado state fair recognized that for what it was an admission to, or that it was a genuine attempt by him to be transparent about the artwork's origins? The program is barely 2 months old, and that merely sounds to me like an attempt by him at creating plausible deniability. So no, still, fuck him.

        8 votes
        1. [2]
          eve
          Link Parent
          Thank you for pointing that out. AI art might be making its rounds online, but what about people who are outside the sphere of its influence? There's a lot of people who don't even know of its...

          Thank you for pointing that out. AI art might be making its rounds online, but what about people who are outside the sphere of its influence? There's a lot of people who don't even know of its existence, or it's a vague notion that might have trickled down to them. This dude was obfuscating and likely on purpose. If he wasn't, he should have in good faith labeled his art as "compiled via AI art creator Midjourney".

          For all the judges knew, Midjourney is just a fancy little art program like photoshop or krita or Sai. It's very frustrating to see and for people to expect the average person to know what the hell Midjourney is.

          6 votes
          1. skybrian
            Link Parent
            If they didn't know before, they do now. I think of it sort of like a hoax to prove a point.

            If they didn't know before, they do now. I think of it sort of like a hoax to prove a point.

            1 vote
    2. [9]
      skybrian
      Link Parent
      It seems like judging art by the effort required is already kind of problematic in the digital art category (where he entered this), where presumably all the Photoshop effects you like are...

      It seems like judging art by the effort required is already kind of problematic in the digital art category (where he entered this), where presumably all the Photoshop effects you like are allowed? If I had to judge which artworks required more effort, I think I'd do badly. It seems like you'd have to be pretty familiar with tools to tell, and maybe some difficult-looking effect is due to a tool you don't use?

      I wonder what the judges' backgrounds are? Glad I'm not a judge.

      9 votes
      1. [2]
        cfabbro
        (edited )
        Link Parent
        I'm sorry, but have you ever actually created any digital artwork yourself, or ever watched someone else do it? Because IMO trying to compare most digital artist's occasional use of filters to...

        I'm sorry, but have you ever actually created any digital artwork yourself, or ever watched someone else do it? Because IMO trying to compare most digital artist's occasional use of filters to augment their work, when the vast majority of the work is still manual input and requires tremendous skill, to the MidJourney generative art process of simply inputting text strings to get a completely finished work as output, is pretty silly.

        I suggest checking out some videos in Twitch's Art category to see just how silly it is... If you can look past how inherently silly all the "sexy" anime girls, and furry art there is, anyways. ;)

        Edit: On second thought, YouTube is probably a better place to go for that, since Twitch's art content is unfortunately pretty overly sexualized. E.g. youtube.com/results?search_query=speed+painting+digital.

        8 votes
        1. rich_27
          Link Parent
          I think what the commenter was trying to point out is that the amount of effort that goes into a piece doesn't actually add value itself. If you were to put a huge amount of effort into cleaning...

          I think what the commenter was trying to point out is that the amount of effort that goes into a piece doesn't actually add value itself. If you were to put a huge amount of effort into cleaning the paintbrush between each stroke, down to the microscopic level, it doesn't mean the painting is going to be that much better as a result.

          If we take a paralysed person who has no use of their limbs, for instance, despite painting solely via holding a brush in their mouth if they create a piece of art that captures the same scene with the same level of detail, the same creativity, and the same ability to evoke emotion as that of an able bodied painter, should we hold that painting as a better work of art?

          I'm totally with you that it really sucks for traditional artists who've spent years honing a craft that someone else can do with seemingly minimal effort due to a new tool, but at the same time I'm sure it absolutely sucked for the people who spent years honing the craft of navigating the ocean by stars and maps when GPS came about.

          The thing I'm excited for is to see what the people who can create fine art using a prompt turn their time and passion towards? If someone could spend hundreds of lifetimes as an artist learning every nuance and developing new ways to express themselves, where would they get to? It feels like we're at the brink of a new era of art, and I'm super excited to see where it will go.

          3 votes
      2. [6]
        cfabbro
        (edited )
        Link Parent
        Rereading my reply to you, I realize I may have sounded harsher towards you than I meant to. Sorry about that. So to answer your questions a bit less dismissively, more seriously, and directly:...

        Rereading my reply to you, I realize I may have sounded harsher towards you than I meant to. Sorry about that. So to answer your questions a bit less dismissively, more seriously, and directly:

        Effects and filters are not usually prohibited in digital art competitions, AFAIK, but there is good reasons for that. Unlike generative art programs, they can't typically be used to create an entire piece of artwork out of thin air. They're instead typically pretty subtly used in the hands of professionals, and more often than not only as a way to simply augment or slightly alter existing elements in a piece that have already been largely crafted by hand. And at most they might be used to create a particular layer within that art, such as the background. Stacking multiple effects and filters can create an entire piece, but even successfully creating a cohesive piece by doing that is itself is a rather unique and impressive skill.

        As for spotting effects and filters, especially when they're being used by amateurs, it's honestly not as difficult as you might imagine once you're familiar with all the tools. So much so that there are even tropes/memes related to their use (e.g. lens flare laser eyes).

        And finally, as for judging art by effort, I never said that should be the sole criteria. Notice how I mentioned "dedication, skill, time, and effort" but even those are not all the things worth considering either. Outcome, and the emotional effect a piece has on its audience, regardless of those other things, is still important too. And that is actually the one few places where generative art can potentially compete with real, human artists. But IMO, without factoring in all those other aspects, that is a pretty hollow way to judge things.

        p.s. BTW, I am not against generative art in the slightest. I think it's really, really cool! But I also think it's remarkably unfair to allow generative art entry into art competitions... unless it has it's own category where it's competing solely against other generative art.

        6 votes
        1. [5]
          skybrian
          Link Parent
          Thanks. Yeah, I haven’t spent time seriously using a paint program since DeluxePaint 2 on the Amiga, and haven’t watched videos of the type you shared either, so I wouldn’t know what to spot and I...

          Thanks. Yeah, I haven’t spent time seriously using a paint program since DeluxePaint 2 on the Amiga, and haven’t watched videos of the type you shared either, so I wouldn’t know what to spot and I might have exaggerated ideas about what you can do nowadays.

          I suppose one way to do a contest might be to have contestants include a video of how they made the artwork? That’s a different art form, though, a combination of teaching and video production, and not everyone is going to want to learn it or be good at it. But it seems there are people who like to do it and others like to watch the art being made, so why not?

          I don’t really expect art to be fair in the sense that some people are more talented than others and can get better results much more quickly and with less practice. You can certainly improve with practice and that can be rewarding, but that doesn’t mean you’ll ever catch up, because they’re improving too.

          The adoption of new tools is going to reshuffle the deck, much like not all silent film actors could make the transition when movies started having sound. I expect there are still going to be people who are more talented at using them appropriately and combining them with other tools, though.

          4 votes
          1. [2]
            cfabbro
            (edited )
            Link Parent
            Eh. Yes and no. Some of the available tools are getting to be remarkably powerful, and are already very difficult to spot the use of, e.g. Photoshop's Content-Aware Fill. And I imagine now that...

            I might have exaggerated ideas about what you can do nowadays

            Eh. Yes and no. Some of the available tools are getting to be remarkably powerful, and are already very difficult to spot the use of, e.g. Photoshop's Content-Aware Fill. And I imagine now that the generative art genie is rapidly coming out of the bottle we will also start to see it being incorporated into more and more digital art programs too. So the line between purely human crafted digital art and generative art is definitely going to get blurrier, and perhaps something more akin to human/AI collaboration (e.g.) will soon even become the norm. As a result, I suspect that deciding where to draw the line as a professional, and how to keep competitions fair, is only going to get harder.

            5 votes
            1. post_below
              Link Parent
              There's a photoshop plugin for Stable Diffusion in beta. And of course there were already impressive AI upscaling, masking and denoising apps out there. You're absolutely right, the line between...

              will also start to see it being incorporated into more and more digital art programs

              There's a photoshop plugin for Stable Diffusion in beta. And of course there were already impressive AI upscaling, masking and denoising apps out there.

              You're absolutely right, the line between human and AI where digital art is concerned is going to get blurry to the point of nonexistent.

              2 votes
          2. [2]
            papasquat
            Link Parent
            Art's not fair in the same way that track and field aren't fair. Genetically, some people are taller than others, some have higher lung capacity, some have better muscle development, some have a...

            I don’t really expect art to be fair in the sense that some people are more talented than others and can get better results much more quickly and with less practice.

            Art's not fair in the same way that track and field aren't fair. Genetically, some people are taller than others, some have higher lung capacity, some have better muscle development, some have a more favorable center of gravity. Socially, some people grew up in a household that emphasized and supported athletic achievement, some people went to schools with good coaches, some people had access to better nutrition. People can't control these things, they just make do with the hands they are dealt, and even though they're not really fair technically, we've accepted that there are very few things we can actually control for to make a level playing field (age, sex, and disabilities being the only three I can really think of). As a society, we've accepted that this is fair enough, and the same goes for the art world.

            What this guy did would be akin to showing up with a Lamborghini on the starting line, and then when he absolutely smokes the competition, people come to his defense saying "Well, he had to spend the money to rent that Lamborghini, and he had to learn how to drive it! Plus, technological progress is inevitable!"

            There are absolutely degrees of fairness, and this is just so wildly outside that, not to mention the fact that he wasn't entirely transparent with the judges on how this piece was generated before he won.

            1 vote
            1. skybrian
              Link Parent
              I'm no art expert, but I think of it more like Duchamp's upside-down urinal. Art is sometimes "deliberately provocative," or let's be real, it's about trolling people. Stunts seem to be allowed if...

              I'm no art expert, but I think of it more like Duchamp's upside-down urinal. Art is sometimes "deliberately provocative," or let's be real, it's about trolling people. Stunts seem to be allowed if you're sufficiently interesting, and getting a Washington Post article written about him seems to be sign that this guy achieved being interesting. It probably won't get a Wikipedia article like the urinal, but maybe it will end up a footnote in some art histories?

              Or consider the conceptual art installations where the artists only writes directions for anyone to execute. Kind of similar?

              I find contemporary art museums are more fun when you accept that there's sometimes a bit of trolling going on and admire things in part for what the artist got away with. (And yep, maybe some of the artists are assholes? Seems like it comes with the territory?)

              1 vote
    3. [14]
      FluffyKittens
      Link Parent
      Counterpoint: the art should hold up on its own merit. Would you be this emotional about someone using a food processor in a pesto competition instead of hand-grinding with mortar and pestle, or...

      Counterpoint: the art should hold up on its own merit.

      Would you be this emotional about someone using a food processor in a pesto competition instead of hand-grinding with mortar and pestle, or using a blowtorch + sous vide on steak instead of grilling? The goal is good food, not how you make the food.

      Neither of those technologies undid the artisanal cooking industry. They make it easier for a rank amateur to compete with a pro chef, but pro chefs have simply added those tools to their kits as well, and yet no one’s mistaking gourmet food made by a pro with gourmet food made by a rando off the street.

      6 votes
      1. [13]
        cfabbro
        (edited )
        Link Parent
        This would be more akin to someone using a Star Trek replicator and submitting that dish to a cooking competition as if it was their own creation. So yes, I would be just as "emotional" and...

        This would be more akin to someone using a Star Trek replicator and submitting that dish to a cooking competition as if it was their own creation. So yes, I would be just as "emotional" and offended were that to take place, since just like in this instance, it would be misrepresenting someone else's hard work as their own, undermining the efforts of those who actually dedicated themselves to the craft, and who had entered the competition in good faith.

        6 votes
        1. [12]
          FluffyKittens
          Link Parent
          Fair enough, I guess it depends on how you view the utility of art. Rhetorical example: is an original Basquiat more artistically valuable than a kindergartener’s doodles if you can’t tell the...

          Fair enough, I guess it depends on how you view the utility of art.

          Rhetorical example: is an original Basquiat more artistically valuable than a kindergartener’s doodles if you can’t tell the difference? Depends on if you deem narrative history and emotional intent to be an inherent part of the artwork itself or not.

          5 votes
          1. [11]
            cfabbro
            (edited )
            Link Parent
            From another of my comments above: So I am not going to begrudge anyone who only cares about utility and focuses exclusively on the outcome more than anything else, at least when it comes to...

            From another of my comments above:

            And finally, as for judging art by effort, I never said that should be the sole criteria. Notice how I mentioned "dedication, skill, time, and effort" but even those are not all the things worth considering either. Outcome, and the emotional effect a piece has on its audience, regardless of those other things, is still important too. And that is actually the one few places where generative art can potentially compete with real, human artists. But IMO, without factoring in all those other aspects, that is a pretty hollow way to judge things.

            p.s. BTW, I am not against generative art in the slightest. I think it's really, really cool! But I also think it's remarkably unfair to allow generative art entry into art competitions... unless it has it's own category where it's competing solely against other generative art.

            So I am not going to begrudge anyone who only cares about utility and focuses exclusively on the outcome more than anything else, at least when it comes to personal enjoyment and appreciation for art. You like what you like for whatever reason, and you do you.

            But the simple fact is that this asshole entered a competition under very questionable pretenses, and by winning it with artwork he didn't actually create himself (in any meaningful sense) he effectively stole the prize money and potential opportunities for career advancement from everyone else. From people who actually did enter in good faith, and put in a significant amount more dedication, skill, time, and effort than him.

            5 votes
            1. [10]
              AugustusFerdinand
              Link Parent
              Say we both make an equally impressive marble sculpture, you do so with a hammer and chisel, I do so with a jackhammer and chainsaw. Is your sculpture somehow better or more "art" than mine?

              Say we both make an equally impressive marble sculpture, you do so with a hammer and chisel, I do so with a jackhammer and chainsaw. Is your sculpture somehow better or more "art" than mine?

              3 votes
              1. [9]
                TemulentTeatotaler
                (edited )
                Link Parent
                Humans care about [human] stories. I'm sure you can write several books on why: humans as expansions of our cognition, some consequence of the evolution of empathy, categorizing entities as...

                Humans care about [human] stories. I'm sure you can write several books on why: humans as expansions of our cognition, some consequence of the evolution of empathy, categorizing entities as agentic/non-agentic, etc. Stories matter, so provenance matters and deception in such is a bad thing.

                @cfabbro clarified that he viewed the deception as the problem, and the use of Midjourney as different from just using another tool/medium. A chainsaw vs a chisel aren't different in the ways a replicator and relic are.

                It's closer to if it was found out Magnus Carlson's legal name was Chadworth and he had smuggled in a computer named Magnus into all of his matches. Since AI can play the best game of chess the audience got to see a better game played. It took some skill to operate the software. No one wants to watch a computer play chess, though.

                Say you have a holy relic that has been visited by countless people for a thousand years in arduous pilgrimages. Is it okay to swap it out with a perfect replica? Are the objects equivalent? The utility of it for the adherents may by the same, and they may appreciate it just as much. IIRC the middle ages had over a dozen claimed "holy grails" that were used as attractions for churches, and many people don't care if the weeping of a statue of the Virgin Mary comes from leaky pipes.

                You can argue that's a victimless crime if it never gets found out. In the case of someone like Mark Landis' donated art forgeries maybe you could even argue that having a dozen holy grails added value and wasn't coming from a bad place or desire to enrich himself.

                One argument against the act of creating a deceptive provenance/history/context/whatever for an object is that the story contains the important information. How would you look at this differently if you were told it was:

                • A self-portrait of an artist as their Alzheimer's progressed. (what it is)
                • From a family member taking care of a parent with cognitive decline.
                • Drawn by a person who lied about having experience with Alzheimer's.
                • A series of images re-fed into an AI with the prompt "this but worse".
                • An AI that was input some sentiment extracted from tweets of old-to-young age groups.

                A painter who spent a dozen years developing skills has changed visual saccades. They've spent years thinking about their style and what it conveys, and if they spend a month on a piece they're submitting to a contest that's a month of human thought and feeling embedded in the work, compared to a hypothetical AI "award winning painting" prompt.

                It's not like an AI is context free. It's still trained on real people, so it's concerning when chatbots quickly become racist or funny when unsupervised learning learns the internet is made of cat videos. It's different from an individual person, though.

                Another argument is that there's always a risk of the real provenance being found and that harms trust on a societal level. One day you catch a partner that was cheating on you in an otherwise side-effect free way in the act. Suddenly the meaning you constructed around years of that relationship collapse and you're infected by doubt. How can you believe anything anyone says, when you have an example of someone reality and claimed reality being indistinguishable? Why would you believe a leaked video of Romney talking about the takers in society when in a dozen years you can get the same video from "video, Romney, political hit piece"?

                A last argument is that beyond the value of the art itself, there's a difference in the reward of the competition going towards the career of an aspiring artist vs. someone using Midjourney. A long while back I shared a small experience about an infomercial that rebranded contact juggling as Fushigi. One of the consequences of the infomercial was performers had people primed to the idea that contact juggling was a trick instead of a skill. The ball did the work, not the years of practice.

                So what would have happened if that killed the profession? New tricks wouldn't be developed to use for the next infomercial. Maybe we'll stop having new "..., Van Gogh, , artstation" prompts to slip in when any graphic designer gets laughed at for wanting to make a living. Or maybe there would be, but with the source of it inverted from generating painter to distributed discriminating audience? That could be better or worse, but it is different.

                7 votes
                1. [8]
                  AugustusFerdinand
                  Link Parent
                  Fakes or fake experiences shouldn't be celebrated, no argument there. But this take that art should require immense effort is flawed. Chisel vs Chainsaw leads to Chisel vs CNC machine leads to...

                  Fakes or fake experiences shouldn't be celebrated, no argument there.

                  But this take that art should require immense effort is flawed. Chisel vs Chainsaw leads to Chisel vs CNC machine leads to Chisel vs AI.
                  If the art is equal in Chisel vs Chainsaw is it still equal if I program a 5-axis CNC to carve it? It's just numbers (coordinates, feeds, speeds) fed into a machine. If telling the CNC what to do is still equal to a Chisel why isn't an AI if I'm the one telling it what to do? If the CNC or AI are somehow less than, then why?
                  Is it because I didn't do it by hand? Is it because the machine is faster? If that's the case then all digital art isn't "art" unless the artist programmed Photoshop from scratch.
                  In the same vein, is photography art? How can it be? The photographer didn't make the camera. All they do is adjust aperture, shutter speed, ISO, and click. The photographer rarely has a story of their own, they're just in the right place at the right time.
                  How about if I'm a quadriplegic. I physically cannot create art, but I have a mind that can imagine wildscapes beyond anyone's dreams. If I feed my ideas into Midjourney is it not art? Should I have hired an artist to paint it exactly how I imagine it? If they did who's art is it then, mine or the person with usable arms?
                  The idea that AI generated art is akin to a Star Trek replicator is false. It's not replicating anything, it's created a new and unique item from inputs, not a copy.

                  Why is architecture considered art if the architect never swings a hammer? They didn't make anything, they put forth an idea and had someone else build it for them.

                  Side by side 99% of people couldn't tell a Cy Twombly from a toddler with mommy's makeup or Basquiat from a kindergartener or a Pollock from someone that puts an M80 in a can of paint. All of these people are "artists" because they have a story (made up or otherwise because who's going to call an artist a liar) about how they felt when making it. Warhol painted a fucking soup can and people want to act like his farts doesn't stink. If I'm in a mood and feed that mood into an AI to get it to produce an image that reflects that mood, how is it not art?

                  No art is context free. It's all derivative of what came before it. The only context free art is from an infant kept in isolation coloring the walls with its own excrement.

                  Another point made above is that it's difficult to make a living making art despite the effort it takes to make it. Which is true. However, there's a question of why. One part is that just because you make something doesn't mean it's desirable or worth anything, regardless of the time spent on it. Deviantart has 61million users, even if we assume only 10% are actually creating anything, that's still 6 million artists churning "tens of thousands of original pieces of art every day". Maybe it's difficult to make a living making art because the market is saturated with artists. Maybe it's saturated because making art isn't that hard.

                  7 votes
                  1. [7]
                    TemulentTeatotaler
                    (edited )
                    Link Parent
                    I don't think art has to require immense effort, I just think the process/context is part of the art and may add value to the finished piece. I remember an early cloud service for brute-forcing...

                    But this take that art should require immense effort is flawed.

                    I don't think art has to require immense effort, I just think the process/context is part of the art and may add value to the finished piece.

                    If the CNC or AI are somehow less than, then why?

                    I remember an early cloud service for brute-forcing password hashes whose FAQ included something like: "Do we guarantee finding the password? No, we guarantee <some high number> combinations will be tried."

                    The idea that AI generated art is akin to a Star Trek replicator is false. It's not replicating anything, it's created a new and unique item from inputs, not a copy.

                    I went to an extreme with the AI prompt being "award winning photo" to try to highlight what is missing. Is that prompt art? What about "awesome space opera"?

                    The end point of that spectrum requires no human thought or commitment. A painter at least spent a bit of time thinking, and they felt the painting was worth spending some number of hours on. The result of their work has that context embedded while the result of the AI does not.

                    The AI may have its own inhuman meaning. Maybe even an arguably richer one. That sort of goes beyond the scope of this reply. When it goes wrong I think its something like Baudrillard's simulacra, where you have copies of copies that increasingly divorce themselves from reality/the source they want to correspond to, eventually becoming a self-referential truth of their own.

                    No art is context free. It's all derivative of what came before it. The only context free art is from an infant kept in isolation coloring the walls with its own excrement.

                    I mean, I'd argue that's a pretty substantial context. Feral children give huge (depressing) insights into the mind. It isn't a question of whether context exists or not (it always does) but what it adds to the story of an object.

                    I'm also definitely not defending Andy Warhol or the art world. I don't know anything about it but the little bits I've heard suggest a bit of it is for tax shelters, pump and dumps, and vanity.

                    Another point made above is that it's difficult to make a living making art despite the effort it takes to make it.

                    That might be cfabbro's(?) but my point was a little different, which was about the ecosystem that these AI are trained in. They rely on centuries of funding artists. If artists stop existing in a post-Midjourney world maybe (definitely doesn't have to be the case) that ecosystem stagnates. e.g., if Midjourney was invented in the paleolithic you wouldn't be able to ask it to give you a Monet.

                    Maybe it's saturated because making art isn't that hard.

                    Honestly nothing is "hard" if we're getting into plausible AGI scenarios. There's nothing special about humans. There's nothing about your thoughts or emotions that aren't eventually subject to the same creeping line of progress that Ken Jenning's mentioned seeing before his loss to Watson. I think the dread of "human's need not apply" is a universal and more interesting conversation than one restricted to art.

                    As a thought experiment, leap forward 4 to 400 years to when we have robust AGI.

                    The AGI plays Cyrano de Bergerac when you meet your future spouse. Anything you can think to say to charm them it can do a lot better. It writes your wedding vows, chooses gifts for anniversaries, names your children. You could step in at any time but it's been demonstrated that will be suboptimal. In fact, you grew up with this AGI, and over billions of instances of it prompting you to make a choice and it giving a better alternative/outcome, you switched it from requiring approval to pending rejection. You go from asking it to make you award winning & heartfelt-looking art or choosing what you want to eat for dinner to being a passive discriminator waiting to tell it "no" or "less peas". But you tell it no less and less because it will never make a worse choice than you; actually it only improves every year. By the time you met your future spouse it chose them, and when they laugh at its jokes it's what's suggested by their AGI.

                    I hope you share the intuition that that's kinda dystopian? If so, where on the spectrum from using Excel or other cognitive aides to make better buildings/art/conversation/etc., to having a strictly-better AI does it happen for you?

                    One solution is what I led with: humans naturally care about human stuff. Call it human chauvinism or tie it in to knowing we have qualia. Sure, the AI is a billion times smarter than use, but we own it, we're keeping it, and we're going to use it to do the dishes and solve climate change while we find fulfillment in being in a jam band or cheering at a friend's concert. We get to endow meaning to activities just because we want to, so we say David Blaine can make a career out of human grit that resonates with us because we know what it feels like to try really hard at something.

                    5 votes
                    1. [2]
                      AugustusFerdinand
                      Link Parent
                      Per the "artist" in this case: Hours upon hours were poured into this piece. Thought, ideas, and iterations were poured into a tool to get a result they wanted. Only here it was prompts instead of...

                      A painter at least spent a bit of time thinking, and they felt the painting was worth spending some number of hours on. The result of their work has that context embedded while the result of the AI does not.

                      Per the "artist" in this case:

                      “I have been exploring a special prompt that I will be publishing at a later date, I have created 100s of images using it, and after many weeks of fine tuning and curating my gens."

                      Hours upon hours were poured into this piece. Thought, ideas, and iterations were poured into a tool to get a result they wanted. Only here it was prompts instead of paint. A keyboard instead of a brush.

                      They rely on centuries of funding artists. If artists stop existing in a post-Midjourney world maybe (definitely doesn't have to be the case) that ecosystem stagnates. e.g., if Midjourney was invented in the paleolithic you wouldn't be able to ask it to give you a Monet.

                      True, but if you ask an artist in the paleolithic era to create something cubist, they wouldn't be able to do so. Since all art is derivative, humans artists are nothing more than biological AI, just as incapable as the AI of creating something truly new. We've just given names to the derivative styles that can all be eventually traced back to our origins. If prehistoric humans didn't slap their hand on a cave and spit dye at it we wouldn't have Monet.

                      Problem with AGI is time travel is impossible. There's no guarantee that the option it gave you is better than the one you'd come up with yourself. You just found the outcome in your specific instance to be an acceptable one. Sure, people are lazy and there are many that will gladly let an AGI make decisions for it as they hurtle toward a WALL-E existence, it takes people on both sides of the bell curve to create that shape. Their lives will look dystopian to some, but utopian to themselves.
                      Machines already do or are at the cusp of doing everything better than humans. Topological Optimization creates structures just as strong as the best human engineers can come up with at a quarter of the required materials and in shapes they couldn't imagine. AI has already begun to write code. Am I a programmer if tomorrow I can type requirements into an AI prompt and it spits out software to do what I want? Why or why not? I gave it the requirements, the idea, it just did the work.

                      I guess the overall question is are artists upset that AI generated art exists or are they upset because it's better than them?

                      6 votes
                      1. TemulentTeatotaler
                        (edited )
                        Link Parent
                        I guess I grouped you in the context of what cfabbro was responding to: ...as arguing that the end product separate from the process was the only important thing. Sorry about that. If the creator...

                        I guess I grouped you in the context of what cfabbro was responding to:

                        Counterpoint: the art should hold up on its own merit.

                        ...as arguing that the end product separate from the process was the only important thing. Sorry about that.

                        If the creator put similar levels of thought into their product and had similar abilities to render those thoughts into a product as an artist that had spent the years they had developing their skills and time spent on a piece I wouldn't devalue using AI. I'd still say it's inappropriate for them to enter a contest if that process wasn't known/accepted.

                        It also isn't inherent to AI generation of art. A painter will always have to spend a significant time developing their skills and more time working on a piece. In the future "award winning art" is probably going to be a viable prompt, and that won't require effort.

                        We've just given names to the derivative styles that can all be eventually traced back to our origins.

                        I get that, and tried to qualify that new styles can emerge from something like procedural/AI generation and an audience serving as a vote of good/bad, just that it is different. TwitchPlaysArt is not the same approach to finding a new style that a Van Gogh is.

                        Going to be my last reply, cheers

                        4 votes
                    2. [4]
                      MimicSquid
                      Link Parent
                      Honestly, I would take a well managed life. Knowing the jokes to tell, always having my meals be ideal, and trusting that the people I meet would be the right people for me. Managing day to day...

                      Honestly, I would take a well managed life. Knowing the jokes to tell, always having my meals be ideal, and trusting that the people I meet would be the right people for me. Managing day to day life is exhausting, and if AGI would handle the parts I don't like, I'd take it.

                      3 votes
                      1. [3]
                        TemulentTeatotaler
                        Link Parent
                        I don't blame you, and I'm pretty far from a luddite. I'm about to have an exhausting and socially fraught week and I've never been great with the social stuff. An appropriate use of technology...

                        I don't blame you, and I'm pretty far from a luddite. I'm about to have an exhausting and socially fraught week and I've never been great with the social stuff. An appropriate use of technology stands to improve most aspects of life.

                        The thought experiment was roughly a gradient experience machine, and at any step of it--when you're in it-- people are probably going to choose to keep using it. It's a higher-order value that it will be kinda heavy on the last day that humans are relevant, before you're posed with the "do I want a better or worse dinner" choices.

                        If you're telling me you wouldn't feel it cheapened the experience a little if you found out something like your partner's wedding vows were taken from Hallmark or a good generator I don't share that intuition.

                        1 vote
                        1. MimicSquid
                          Link Parent
                          I mean, our wedding vows were pulled from a long tradition that sets pretty narrow guidelines for the specific things to say. We may have picked the particular words, but it was with the influence...

                          I mean, our wedding vows were pulled from a long tradition that sets pretty narrow guidelines for the specific things to say. We may have picked the particular words, but it was with the influence of our shared history and culture guiding our pick. Under the hood would it be that different if we'd gone to a computer and had it generate the top three picks popular among our age range and culture? Only to the degree that it made clear how narrow our choices really were, given our demography.

                          2 votes
                        2. rich_27
                          Link Parent
                          I've been reading through this thread and found your discussion with AugustusFerdinand and MimicSquid fascinating. I think there was one thing you missed in your analysis, and that is the...

                          I've been reading through this thread and found your discussion with AugustusFerdinand
                          and MimicSquid fascinating. I think there was one thing you missed in your analysis, and that is the discerning eye it takes to select from the output of a generator. Sure, maybe infinite monkeys can write Shakespeare, but it is all for naught unless you have someone able to recognise it and select that as the output of the system. There is a huge difference between someone who writes an intricate prompt and takes the first piece the generator creates and submits it to the competition, and someone who writes such a prompt but keeps at it until they have a result they think is truly award winning. Even if we get to a point where AI can generate something incredible just from "award winning art", there is still a huge amount of value added by the person who flips through that output and selects the piece to submit. All in all, a relatively minor point, but I think an important one to add to the discussion.

                          2 votes
  3. [8]
    Bearskin
    (edited )
    Link
    "Of course that was going to happen." That was my initial thought regarding the art competition, and I'm glad that this has sparked a lively discussion of the topic here and elsewhere. I think I'm...

    "Of course that was going to happen." That was my initial thought regarding the art competition, and I'm glad that this has sparked a lively discussion of the topic here and elsewhere. I think I'm chiming in just to vent because this stuff keeps me awake at night, and I'm not seeing a lot of talk around the potential to weaponize this technology. Deepfakes raised concerns, sure, but I think it's only a matter of time before this is used to destabilize parts of the world through believable propaganda.

    I'm far from an expert on this subject, and just played around with the simple A.I. image generators for a few days before moving on. Using DALL·E 2 was a different matter entirely. People, landscapes, interiors, etc. My jaw hit the floor seeing how far this technology advanced in such a short amount of time. OpenAI discourages the use of real people, and seems to prevent certain images from being generated.

    But how long before this cat is out of the bag?

    If Putin releases a seemingly real video of Volodymyr Zelenskyy's surrender--what happens? If a video of Joe Biden comes out before the next election appearing to discuss stealing the election--seated next to George Soros and Hillary Clinton as they hatch a diabolical plan to throw Republicans in prison on false charges--what happens? It seems laughable, but a certain percentage of people around the globe are going to believe this stuff is reality. Throwing random prompts into DALL·E 2 was fun for a few days--until it became terrifying to think about what someone with bad intentions could do with it.

    9 votes
    1. [3]
      cfabbro
      (edited )
      Link Parent
      FYI, that exact thing was already tried and thankfully debunked pretty swiftly: Deepfake video of Zelenskyy could be 'tip of the iceberg' in info war, experts warn But IMO you're not wrong to be...

      If Putin releases a seemingly real video of Volodymyr Zelenskyy's surrender--what happens?

      FYI, that exact thing was already tried and thankfully debunked pretty swiftly:
      Deepfake video of Zelenskyy could be 'tip of the iceberg' in info war, experts warn

      But IMO you're not wrong to be concerned. Imagine something like that being created and released shortly after the death of such a public figure, so there is no way for the target to directly deny it? That would be a lot harder to debunk, might be believed by a lot more people as a result, and so end up being far more effective and harmful. Misinformation is already difficult enough to combat, but sadly it looks like it's only going to get much much harder in the future due to all these machine learning tools being developed without much consideration of their potential for misuse. :(

      8 votes
      1. [2]
        kfwyre
        Link Parent
        To add on to this, there’s also the idea that malicious use of these tools could easily be deployed by and against everyday people who don’t draw sufficient media scrutiny for adequate debunking....

        To add on to this, there’s also the idea that malicious use of these tools could easily be deployed by and against everyday people who don’t draw sufficient media scrutiny for adequate debunking.

        A jealous ex could ruin their former partner’s new relationship with a smattering of fake yet very convincing “evidence”. Someone could get a colleague fired at work by fabricating a set of very compromising material on them.

        Faking these sorts of things is technically possible already, but it still requires tons of work and technical knowhow. Soon, though, these tools will make the barrier to entry very low, yet they will produce convincing output at a very high level. Furthermore, it’s likely that many (if not most?) malicious uses will occur at levels below or outside of expert attention.

        5 votes
        1. scissortail
          Link Parent
          It all makes me wonder if we will eventually just stop trusting digital images and video altogether. It would be strange times if we have to go back to film to even approach a baseline for...

          It all makes me wonder if we will eventually just stop trusting digital images and video altogether. It would be strange times if we have to go back to film to even approach a baseline for trustworthiness.

          6 votes
    2. [2]
      kfwyre
      Link Parent
      I feel the same way. My stomach also turns when I think about the capability for just truly horrible imagery. As far as I understand it, most of these image generation programs have some built in...

      I feel the same way.

      My stomach also turns when I think about the capability for just truly horrible imagery. As far as I understand it, most of these image generation programs have some built in guardrails to prevent people from using them to produce grotesquely violent or pornographic images, for example. I don’t think it’s that long before people either learn to escape those guardrails, or just develop a new tool without the same scruples.

      Growing up in the “golden” age of internet shock sites, I’ve seen some scarring images that still linger with me to this day — decades later. The idea that these tools could effortlessly pump out a flood of images (and eventually videos) specifically tailored to maximize shock value across every conceivable axis? It makes me queasy — especially when I consider them venturing past the realm of “shock” and into truly offensive or abhorrent imagery.

      3 votes
      1. stu2b50
        Link Parent
        For better or for worse this is already the case. Stable Diffusion is a version of these with comparable performance to DALL-E or Midjourney (they all have their ups and downs), but is entirely...

        As far as I understand it, most of these image generation programs have some built in guardrails to prevent people from using them to produce grotesquely violent or pornographic images, for example. I don’t think it’s that long before people either learn to escape those guardrails, or just develop a new tool without the same scruples.

        For better or for worse this is already the case. Stable Diffusion is a version of these with comparable performance to DALL-E or Midjourney (they all have their ups and downs), but is entirely open source. You can download it right now. And, of course, you can do whatever you want with it. Most of the "guardrails" are hard-coded, after all. I believe some distributions of Stable Diffusion have a nudity filter that returns an image of Rick Ashley for prompts that heuristically will lead to adult images, but you can just comment it out (this hackernews thread involved people discussing how to turn the filter off because it had false positives: https://news.ycombinator.com/item?id=32679772)

        https://news.ycombinator.com/item?id=32572770 this is a hackernews thread about someone who built a porn generator from Stable Diffusion (linking to the hackernews thread since it has more context - also, the website itself immediately has images of naked women, so NSFW)

        These are pretty fun to work with so there's a huge amount of interest and development in Stable Diffusion, so I would anticipate it to advance quickly as well under the open source community.

        8 votes
    3. [2]
      Fiachra
      Link Parent
      While watching the imminent financial ruin of Alex Jones unfold, I let myself feel optimistic for a while that the post-facts era would eventually end when society adapts to deal with rampant...

      While watching the imminent financial ruin of Alex Jones unfold, I let myself feel optimistic for a while that the post-facts era would eventually end when society adapts to deal with rampant misinformation. Until I remembered that AI-generated images exist. It seems pretty clear to me that the post-fact era is only getting started, and media literacy is going to be a vital skill to teach to everyone, especially children, if we're going to get through it.

      3 votes
      1. Bearskin
        Link Parent
        I have experienced some schadenfreude while keeping up with the Alex Jones situation, but would feel more optimistic if the dark money pipeline was exposed and dismantled. You can de-platform...

        I have experienced some schadenfreude while keeping up with the Alex Jones situation, but would feel more optimistic if the dark money pipeline was exposed and dismantled. You can de-platform Jones, remove the Infowars brand, shut down the LLC, etc. He's going to be right back in the game after a few anonymous donations. Throwing $50 million at someone like Jones to destabilize a country is more cost effective than a single F-16 fighter jet at roughly the same price.

        4 votes
  4. [5]
    post_below
    Link
    It was only a matter of time. It's remarkable how fast this is all moving, Stable Diffusion was released about a week ago.

    It was only a matter of time. It's remarkable how fast this is all moving, Stable Diffusion was released about a week ago.

    7 votes
    1. [4]
      skybrian
      Link Parent
      This was using MidJourney, which has been out for a few months. There are probably more reporters looking into it now, though. The Discord message wasn't hard to find because a lot of people...

      This was using MidJourney, which has been out for a few months. There are probably more reporters looking into it now, though.

      The Discord message wasn't hard to find because a lot of people upvoted it. I was able to get a better look at the images.

      From the article:

      Allen said he had clearly labeled his submission to the state fair as “Jason Allen via Midjourney,” and once again noted the human element required to produce the work.

      Giving it its own category in the art competition should help.

      7 votes
      1. [3]
        post_below
        (edited )
        Link Parent
        Midjourney reportedly incorporated learning code from the Stable Diffusion source, which improved it's quality significantly, that's why I mentioned it. Maybe this guy's submission was created...

        Midjourney reportedly incorporated learning code from the Stable Diffusion source, which improved it's quality significantly, that's why I mentioned it.

        Maybe this guy's submission was created before the change, but either way the speed that AI image generation is changing the landscape is mind blowing.

        Edit: Just googled, Midjourney went into open beta on July 12, 2022.

        2 votes
        1. [2]
          skybrian
          Link Parent
          MidJourney has announced and deployed some quality improvements, so it’s a moving target. I haven’t really noticed the changes, but my testing is limited. It still seems overall weaker than DALL-E...

          MidJourney has announced and deployed some quality improvements, so it’s a moving target. I haven’t really noticed the changes, but my testing is limited.

          It still seems overall weaker than DALL-E and dreamstudio.ai for the few queries I’ve tried, but that can vary a lot; they all have their strengths.

          Even with the sharing of some algorithmic improvements, they are starting from different datasets and training for different amounts of time, so I expect they’ll remain different.

          2 votes
          1. post_below
            Link Parent
            In my limited experience DALL·E is the best of the bunch but Stable Diffusion (dreamstudio being the official web app for SD) seems to be catching up. Their release of img2img was big, and the...

            In my limited experience DALL·E is the best of the bunch but Stable Diffusion (dreamstudio being the official web app for SD) seems to be catching up. Their release of img2img was big, and the open source community will keep doing amazing things with SD.

            There are already SD projects out that can run it locally on a PC with only 6 gigs of VRAM!

            It's times like this when the power of open source reminds me of the early internet, when all the cool toys were wide open to anyone with the time to spare.

            3 votes
  5. [2]
    stu2b50
    Link
    One funny hypothetical I heard related to this is that we'll start training nn models to detect nn generated art. ...but authors of the generating neural network can use the neural networks that...

    One funny hypothetical I heard related to this is that we'll start training nn models to detect nn generated art.

    ...but authors of the generating neural network can use the neural networks that detect it to further improve their neural networks ability to create undetectable art...

    ...which basically makes a Generative Adversarial Network but split in the physical world.

    (For context, GANs are a type of neural network that involve two networks that train adversarially - the generator, and the discriminator. The generator is trained to generate output that looks like the training set, by training against the weights of the discriminator which at the same time is trained to tell what is the generator and what is the training set)

    6 votes
    1. DataWraith
      (edited )
      Link Parent
      Tangent: That could be done using Adversarial examples. The idea is that you modify your image imperceptibly (to a human), e.g. by modifying the hue of a single pixel, but the machine is misled...

      Tangent: That could be done using Adversarial examples. The idea is that you modify your image imperceptibly (to a human), e.g. by modifying the hue of a single pixel, but the machine is misled completely into classifying, say, an image of a school bus as an ostrich.

      The same principle would apply to neural networks trying to detect generated art, and there's probably not even a need to re-train the image-generating NN, because you could make arbitrary images pass the test using externally applied adversarial perturbations.

      Of course, given that there is interest in generating adversarial examples, there is also a lot of research into making networks more robust to adversarial perturbations... so I can't predict who is going to win this particular arms race.

      2 votes
  6. skybrian
    Link
    He used AI to win a fine-arts competition. Was it cheating? (Washington Post) [...] [...] So Allen made some changes to MidJourney's output. For my purposes, that's cheating in the opposite...

    He used AI to win a fine-arts competition. Was it cheating? (Washington Post)

    The pieces that really caught his attention, though, were what he now calls his “space opera theater” series. He started with a simple mental image — “a woman in a Victorian frilly dress, wearing a space helmet” — and kept fine-tuning the prompts, “using tests to really make an epic scene, like out of a dream.” He said he spent 80 hours making more than 900 iterations of the art, adding words like “opulent” and “lavish” to fine tune its tone and feel. He declined to share the full series of words he used to create his art, saying it is his artistic product, and that he intends to publish it later. “If there’s one thing you can take ownership of, it’s your prompt,” he said.

    [...]

    When he found images he really liked, he pulled them into Adobe Photoshop to remove visual artifacts. In one image, the central figure was missing a head, so he also painted in a crop of dark, wavy hair. He used another machine-learning tool, Gigapixel AI, to increase the photos’ quality and sharpness, then printed the three pieces on canvas — all variations on the French phrase for “space opera theater,” which he thought sounded cool — and drove to submit them to the state fair.

    [...]

    One of the judges, Dagny McKinley, an author and art historian who runs a playwright festival in nearby Steamboat Springs, remembers walking past Allen’s canvas and being immediately drawn to a piece that felt reminiscent of Renaissance art.

    “It had an immediate story: People looking out into another world, everyone with their backs to you, no one facing or engaging with the viewers,” she said. “You get interested: What are they seeing?”

    McKinley said she did not realize the art was AI-generated but said it wouldn’t have changed her judgment anyway; Allen, she said, “had a concept and a vision he brought to reality, and it’s really a beautiful piece.”

    So Allen made some changes to MidJourney's output. For my purposes, that's cheating in the opposite direction, because I make images to judge how good the AI generator is. (When I share images I will often share the prompt and say how many candidates I'm choosing from, like "best of four" or "best of eight.")

    One difference I've seen between MidJourney and DALL-E is that DALL-E has a better idea of how big things are supposed to be and is a bit reluctant to have things be the wrong size, while MidJourney is more likely to have things of wildly different sizes, giving the images a more otherworldly feel. You can see that in one of Allen's images where there is a tiny figure, a giant woman, and three mid-sized women standing on the same platform.

    4 votes
  7. MimicSquid
    Link
    An artist/writer duo chime in on this debate (in joking fashion.)

    An artist/writer duo chime in on this debate (in joking fashion.)

    1 vote
  8. tealblue
    (edited )
    Link
    I think people get too caught up in the moralization of if AI is good or in the argumentation of if the work that AI produces is actually all that great. On the whole it's important to keep in...

    I think people get too caught up in the moralization of if AI is good or in the argumentation of if the work that AI produces is actually all that great.

    On the whole it's important to keep in mind that AI, unlike previous technological developments, won't just displace humans; it has the ability to replace all forms of human work unless we support the arbitrary but valid paradigm that "AI is bad". Another way to think about it if that sounds too dogmatic is that we should let the market price in people's taste against AI and let governments form policies accommodating people's ethical stances against AI.