Codified Likeness Utility

November 26, 2022 6:15 AM

Using the Midjourney AI, graphic designer Johnny Darrell has come up with some mind-bending imaginings [Facebook link] of a "Tron" directed by Alejandro Jodorowsky.
posted by ryanshepard (27 comments total)

Alternative non-Facebook links: 1, 2.
posted by urbanwhaleshark at 6:33 AM on November 26, 2022 [5 favorites]

It's such a pleasure to see someone who understands drama and design exploring the worlds these new tools open up.
posted by drdanger at 7:43 AM on November 26, 2022

This is so incredible. What kind of level of detail is the inputs does one need to use to achieve results like this, I wonder?
posted by ukdanae at 7:45 AM on November 26, 2022 [2 favorites]

The thought that in a generation or two, this kind of automation will be able to create whole movies this way (maybe not great movies, but no worse than bad movies made by humans) is mind-blowing.
posted by rikschell at 8:11 AM on November 26, 2022 [1 favorite]

This is so incredible. What kind of level of detail is the inputs does one need to use to achieve results like this, I wonder?

Comments by Darrell from the Facebook thread:

"My prompts were simple variations of "production still from 1976 of Alejandro Jodorowsky's TRON --ar 3:2" and then added in extra details like film lenses and "light cycle" or "disk" or "Recognizer" etc."

"I basically asked for a production still of the movie by Jordo's Tron and gave some basic analog film data like ISO ratings etc. MJ did the rest! No remixing or Photoshop (other than adding the text/title to the movie poster). No image prompts."

The thought that in a generation or two, this kind of automation will be able to create whole movies this way (maybe not great movies, but no worse than bad movies made by humans) is mind-blowing.

The best outcomes I've seen from Midjourney are ones where it was used for design ideas that were then rendered non-digitally (some v. cool Halloween costumes this year, for instance). I can see the back and forth between AI and human having some very interesting possibilities - maybe better than either could arrive at independently.
posted by ryanshepard at 8:15 AM on November 26, 2022 [1 favorite]

::inhales::

ALL AI GENERATIVE ART IS ART THEFT!

::exhales::
posted by Faintdreams at 8:21 AM on November 26, 2022 [10 favorites]

Yes, you're right about art theft. We shouldn't even call it AI, it's machine... I wouldn't even call it learning so much as training. If we could ever invent a system that truly was AI, it wouldn't be theft any more than human art is, but training computers on unlicensed artwork produces very cool, very unethical results.
posted by rikschell at 8:26 AM on November 26, 2022 [2 favorites]

My favorite of the Midjourney-inspired Halloween costumes is here (scroll about halfway down), BTW, courtesy of Chris Brown's excellent Field Notes newsletter.
posted by ryanshepard at 8:30 AM on November 26, 2022 [1 favorite]

Well, all art is theft. AI art is mechanized theft though, and a lot faster.

I’d like better arguments than “theft” personally.
posted by seanmpuckett at 9:22 AM on November 26, 2022 [7 favorites]

An artist will gladly tell you their inspiration, whose work they admire, whose work you might enjoy. The machine as designed does none of these things. It steals with no possibility of attribution.
posted by Zalzidrax at 9:35 AM on November 26, 2022 [5 favorites]

I know it's generally considered uncool to break containment on this kind of thing but I have to wonder what Jodorowsky thinks/would think about these images. It's an interesting question in that if Jodorowsky were to come up with concept art for Tron today, it would of course look very different from what he would have come up with in 1976.
posted by gentlyepigrams at 9:45 AM on November 26, 2022

I’d like better arguments than “theft” personally.

There are none. It's just a reactive argument, the same one people have been trotting out whenever there's some new technological innovation.
posted by iamck at 9:54 AM on November 26, 2022 [1 favorite]

I think that’s unfair. These algorithms are presumably being trained on artwork by artists who have not given permission for their work to be used in such fashion. “Theft” is generally the wrong word for intellectual property violations, but there’s definitely something unethical about how these machines are being trained on the unpaid labor of others.
posted by rikschell at 10:19 AM on November 26, 2022 [6 favorites]

Speaking as an artist there's no way I could tell you all the influences in my work. My practice is built up over a lifetime of observing ("stealing") other sources. I can only tell you what my most deliberate influences were when working on a particular piece. Same as with the machine.

Just about all these conversations about AI art show a lot of anxiety about money and copyright and a lot less interest in Art and what it's good for, and that seems incredibly dismal to me.

If tomorrow the next generation of AI diffusion came out and the public found it met all their art needs* and every artist was out of work, I'd say "A job well done".

That asterisk is why I wish these conversations were better. What are those art needs, and how can they be met? Clearly AI cannot do it now, can it ever? Will this change the narratives in art, and why not in the opposite way? This is a real litmus test and crucible for art, the like of which we may not have seen since photography or even before.

Same with github copilot. Generally speaking it seems to be a clear public good. It has some issues people mostly aren't talking about but... The question people are asking of it is: "is it breaking copyright" and we understand that this really means "my finances are part of a system that funds progress by making new progress temporarily unequally distributed, and uses the resulting differential to generate power". Copyright is part of the engine of capitalism in other words.

So what is wrong with the de-financialization of this era of creation? What system can result?

I don't know, I'm not the brightest I just know a bit about art and AI. But I wish I could find the forum where people are talking about these things seriously. DM me if you know.
posted by tychotesla at 11:09 AM on November 26, 2022 [37 favorites]

Thanks for your insight tychotesla. AI art discussions tend to take the stridency you'd see more often when arguing about, say, religion rather than philosophy for some reason. But I'm not trying to make my living as an illustrator so it's easier for me to say so.
posted by gwint at 11:40 AM on November 26, 2022 [1 favorite]

I've been using Stable Diffusion to create imagery for a table top roleplaying game (Apocalypse World). (Here's the slide deck of 'good stuff.') SD radically lowers the bar for creating usable imagery.

It definitely still has trouble when you push at the bounds of the kinds of things found in the training set or ask for something /very/ specific; but I also haven't been messing with img2img yet, which is where I think one goes to get cleaned up images of weird stuff.

So, I think this all points to the main problem with human art: It's way too expensive for most users, both in terms of money and time. Iteration with AI tools is extremely cheap; just adjust your prompt and pick a new seed, get new results a few seconds later. Over time, you learn what prompts are more likely to give nice results... Just like working with a human artist, but much faster.

Speed and low cost ALWAYS open up undreamt-of use cases, and this ends up being great for most people. The effects on the industry are going to be more complicated, but will prooooobably end up looking like further centralization with a few winners and tool providers making most of the actual money. Ironically, this already describes the art world, since the invention of photography displaced mass-market portraiture.

We're going to continue living in a world where it becomes much easier to create high-enough quality content. This is basically where we've been with music for the last twenty years and books for the last ??? years: There's more high quality content than anyone can consume. I would argue that this is the really essential 'phase' of a creative market: do you have a few high quality things that everybody has read or an ocean of good-to-high quality stuff where you can find the stuff that matters to you and ignore everything else?

We're finding our way into a post-scarcity world with respect to human knowledge and creativity. But the physical world still has plenty of scarcity, or at least enough to keep capitalism rolling.
posted by kaibutsu at 12:15 PM on November 26, 2022 [6 favorites]

My current feel with AI imagery is that it’s an accessibility tool.

It’s like now there are “elevators” for art where before you had to climb a shitload of stairs or pay someone to carry you. What is bad about allowing people unable to wield a brush or pen to turn their thoughts into a visual representation.

We will always need human artists to push the boundaries, create their own realms and, for others, realize exacting requirements.

For everyone else, it is a marvel of empowerment.
posted by seanmpuckett at 12:56 PM on November 26, 2022 [4 favorites]

Aside from the misappropriation argument, which is certainly worth being upset about*, there's another argument that notices that if you look at enough of it, AI art tends to be boring and melty suffer from a lack of intentionality. Note that all of those things can be fixed by careful editing, careful prompt creation, or just having it generate a bajillion images and picking ones that happen to look nuanced and intentional and not as melty.

I worry about commercial artists, licensing work to uncaring corporations, and the speed at which AI makes images is upsetting, but I am less worried about fine art. AI can't do traditional media yet so there's still time to not worry, but beyond that, I think the reason people buy art is they are looking for a human connection, and they like to see the intention of the artist. You can see the hand of the painter. You can look at the marks and know they were laid down by another person.

I can make a painting, and it's good or bad as far as how accurately it represents reality or what my goals were with it, but that goodness or badness doesn't necessarily translate into sales, because what matters really is the connection the painting makes with the buyer. So my first sales went to people I know (because they already had a connection to me) and I do okay with communities I am part of, and if I were any good at business at all(which I am not), a large part of my artistic practice would be dedicated to becoming part of lucrative communities.

Now that I think about it, a lot of commercial art exists to provide that warmth and connection. Sometimes I like to do a deranged rant about how photography killed painting dead over 150 years ago, and every photographic advance since then has only served to blow pieces off the shambling corpse. But there are still commercial artists (more than ever, many less well paid than ever), and I think when someone hires, say, a magazine illustrator over taking a snapshot or buying some stock art, they are looking for that emotional connection. (Similarly, I think one of the things that distinguishes an art photographer from, say, a snapshot taker is the ability to make that connection.)

The thing about the art of Jodorowsky's Dune is that it was Jean Giraud, Chris Foss, and H.R. Giger. Those are all artists whose work I admire. Jodorowsky himself is relentlessly human. Objective amazingness of the images aside, it is the messy human story of it that I admire. If I had a copy of the artbook, I wouldn't show it to people and expect them to be impressed but its mere physical existence, I would be like "here's a story..."

IMO AI generated Jodorowsky's Tron is perverse and extremely wrong. Which, I guess, is why it was worth doing, and why it warrants a post.
posted by surlyben at 1:00 PM on November 26, 2022 [2 favorites]

Some really good takes here in this thread, everyone is going to (can only!) approach this from their own perspective. In my own profession, finance and accounting, a single cheap laptop computer today could do the work of 200 workers from 1970. Rather than rendering them obsolete, this advancement empowered finance professionals - they are still commonplace in every organization on the planet. Technology didn't replace them, it actually frees them from doing 99% of drudge work to focus on what truly makes a great finance professional or accountant - astute judgement, creativity, communication skills. Those same core skills - are alluded here in this thread, which separates the mass of "melty" and incoherent AI-generated art from a truly great piece of art which communicates something special between the creator and their audience.

I've heard it said the difference between photography and painting is that photography is subtraction - you are in the world, and you have to choose what to exclude to make a good photo. While painting is addition - you start with a blank canvas, and you have to choose what to include. Ultimately, AI-assisted art may end up a combination of both - the ability to sift through thousands of generated ideas and images, and then using that to build up a piece that your audience can connect to.

Regarding fine art, this piece made the news some months back winning for first place in a state fine art competition. If this is all that AI-generated art ever is capable of (like the continued promises of AI self driving cars) I suspect interest will quickly wane - many pieces look "samey" after awhile, and people will eventually go, ah this was generated by an AI, not a real human, still, very pretty. But if it continues to evolve and improve, the future of art will get very interesting...
posted by xdvesper at 2:10 PM on November 26, 2022 [3 favorites]

Beyond questions of money, employment and intellectual property theft, I think there's something existentially terrifying about outsourcing humanity's creativity to machines. I always assumed it would be the drudgery that got automated away, leaving people to focus on higher pursuits like art. Instead, it's art that's being taken over by machines. If things keep accelerating like they are, machine learning algorithms could studiously avoid training their data on copyrighted content and still render whole swaths of artists obsolete. But beyond even that, how much of our thinking are we going to end up outsourcing to algorithms? And what kind of world will we create by doing that?

Some of the art being created by AI is truly amazing and visionary. That's what worries me.
posted by ThisIsAThrowaway at 3:28 PM on November 26, 2022 [1 favorite]

These images feel like they came out of a mix of Moebius and 1970s sf. Gorgeous.
posted by doctornemo at 4:50 PM on November 26, 2022

> But I wish I could find the forum where people are talking about these things seriously.

here and here?
posted by kliuless at 9:06 PM on November 26, 2022 [1 favorite]

These images are beautiful, and I think they really speak to AI art's primary strength, which is being able to combine mediums and visual styles to create striking novel forms. To replicate these images conventionally he would have had to spend weeks in photoshop, or build a set and some costumes and pay people to pose for him. It probably never would have happened, but it's a really cool idea, and I'm glad I got to see it.

Even if AI art is only the seed of a new idea, to brainstorm, it is still incredibly valuable. It allows the user to just enter a phrase like an incantation and summon an image into being. 99% of the results are complete garbage, but the only real limitation is how well you can describe what you want to see in a way the AI will understand you and how much patience you have to retry. What would a velociraptor look like in a mech suit? Garfield in a painting by Bruegel? Find out!

The thought that in a generation or two, this kind of automation will be able to create whole movies this way (maybe not great movies, but no worse than bad movies made by humans) is mind-blowing.

It's already starting. Cat music visualizer, Voyage Through Time. You can run it as an overlay on video (WarpDiffusion). It will only get more controllable with less artifacts as time goes on, and with more tools to create consistent characters it will almost certainly lead to amateur directors in a fairly short period of time. Definitely less time than a generation or two. It's an incredibly fast-moving technology right now.

I joined MidJourney after reading This post on a comic Ursula Vernon made.

Conventional art takes time. It takes time to learn, it takes time to make, and not all of the works you create will be good ones. There's a lot of tailings in the pile when you go looking for gold. Like Ursula said in that Twitter thread, the alternative to AI art there wasn't hand drawing buildings, it is not doing the comic at all, because the cost in time and tedium is too high. Nobody has infinite free time, and a lot of amazing ideas are not profitable, or too much of a risk to gamble weeks on. If you can jumpstart that process and start with a seed of something "good", you can use your skills to make it "great" in half the time, and that is a labor saving improvement.

It can enable people to focus more on the soul and the meaning, which are what we care about in art to begin with. I find the comparisons to photography apt. Does it matter whether the image came from an oil painter or a camera? Why? Are things that take longer to produce inherently more valuable? How much? What is art for?

There are limitations. AI art is great at reproducing style but blind to substance, it doesn't understand what details are important or why, which is why everything looks melty and samey, it's basically an average. It's chock full of body horror as a result. It will frequently produce works that are aesthetically pleasing but completely soulless, creating shape with no meaning. The soul is missing, but why? With conventional art, every stroke of the brush is defined by intention, but with AI there is no intent. These are the things I wish we could talk about because they're so much more interesting & I agree with tychotesla. There's definitely a conversation to be had about compensating artists for including their work in a training set, right to opt out your work, etc, but it shouldn't be the only conversation.

It’s like now there are “elevators” for art where before you had to climb a shitload of stairs or pay someone to carry you. What is bad about allowing people unable to wield a brush or pen to turn their thoughts into a visual representation.

This. The only way that it makes sense to vilify AI art as a whole is if people think the ideas of folks who don't have the time, ability, or drive to dedicate literal years of their life to the craft shouldn't get to say anything. It's the equivalent of denigrating packaged dinners because everyone should cook from scratch or pay someone to cook for them. The guy that works 2 shifts at McDonalds and dreams of a better world shouldn't get to show it to others? It has to live in his head forever, unexpressed? We need more range of ideas, from more sources, not less. Creativity is not a zero-sum game. Is there no room for joy among the cynicism?

We're finding our way into a post-scarcity world with respect to human knowledge and creativity.

Yes! What makes art "good enough" for your purpose? Are you so invested in your tabletop game with your friends that you are willing to pay a working artist $200 for their time to depict your characters, or does it make more sense to reroll some prompts like "proud male elf, blonde braid hair, green eyes, illustration in the style of mucha and dragonage, forest background, headshot --test" until you find something that you can relate to? (Hell, take that image to a working artist as a starting point and get some phenomenal results!) We are on the cusp of having completely customized art tailored to extremely niche interests. Want to take pictures of your kid and make art of her as the magic princess she's obsessed with? Go for it! (DreamBooth)

I've been using Character AI to basically make my own personal interactive fanfiction. I'm not a writer, but the AI is advanced enough that I can make some great stories by lobbing ideas back and forth, it's a writing partner. I can take descriptions it produces into MJ and create icons for the characters, or for scenes the CAI describes to better visualize them. I can easily imagine a future where these are streamlined into video games, where an NPC that you talk to remembers what you did together and has a coherent personality. Where Hatsune Miko or whatever you're into reads you a bedtime story. Personally, I don't feel my interests are well-served by the mass media juggernaut that wants me to watch endless Marvel movies. I'm excited to see where these technologies go, because I think they can lead to more niche things I'll enjoy.
posted by Feyala at 10:09 PM on November 26, 2022 [5 favorites]

Narratives - "To the extent that large language models (and I should note that while I'm focusing on image generation, there are a whole host of companies working on text output as well) are dependent not on carefully curated data, but rather on the Internet itself, is the extent to which AI will be democratized, for better or worse."

[MidJourney’s David Holz:] “I had this goal, which was we needed to somehow create a more imaginative world. I mean, one of the biggest risks in the world I think is a collapse in belief, a belief in ourselves, a belief in the future. And part of that I think comes from a lack of imagination, a lack of imagination of what we can be, lack of imagination of what the future can be. And so this imagination thing I think is an important pillar of something that we need in the world. And I was thinking about this and I saw this, I’m like, ‘I can turn this into a force that can expand the imagination of the human species.’[1,2] ...

“So that was kind of the vision. But I mean, there is a lot of stuff we didn’t know. We didn’t know, how do people interact with this? What do they actually want out of it? What is the social thing? What is that? And there’s a lot of things. What are the mechanisms? What are the interfaces? What are the components that you build this experiences through? And so we kind of just have to go into that without too many opinions and just try things. And I kind of used a lot of lessons from Leap here, which was that instead of trying to go in and design a whole experience out of nothing, presupposing that you can somehow see 10 steps into the future, just make a bunch of things and see what’s cool and what people like.”

...

Narratives are tempting but too often they are wrong, a diversion, or based on theory without any tether to reality. Narratives that are right, on the other hand, follow from products, which means that if you want to control the narrative in the long run, you have to build the product first, whether that be a software product, a publication, or a company.

That does leave open the question of Musk, and the way he seemed to meme Tesla into existence, while building a rocket ship on the side. I suspect the distinction is that both companies are rooted in the physical world: physics has a wonderful grounding effect on the most fantastical of narratives. Digital services like Twitter, though, built as they are on infinitely malleable software, are ultimately about people and how they interact with each other. The paradox is that this makes narratives that much more alluring, even — especially! — if they are wrong.

posted by kliuless at 10:49 PM on November 26, 2022 [1 favorite]

These are great! I didn't realize that AI art could do stuff like this.
posted by Hicksu at 12:04 PM on November 27, 2022

FYI the AI art generators running on LAION (which the most popular ones do) has stuff like private medical photos without consent, among the countless copyrighted art. They used royalty-free music for the music AI equivalent, but not for the images, because they determined artists don't have enough financial and legal power to make a dent, whereas the music industry does. So their decisions were entirely cold and capitalistic, and I can't support it in good conscience because of that.

I am open to new art forms but not ones that exploit other people's privacy and work.
posted by picklenickle at 8:05 AM on November 29, 2022 [1 favorite]

To all the artists who are (rightfully) suspicious of the ethics of the training sets used for AI art, Mat Dryhurst and Holly Herndon created a tool to see whether your work has been used in a training set, and an option to opt out of being included in future training sets: https://haveibeentrained.com/
posted by Cpt. The Mango at 1:01 PM on November 29, 2022 [1 favorite]