The tools of generative art are already in the hands of people who would use them to process large amount of copyrighted work without compensating the artists. I ended
the first Midjourney post with an analogy, comparing them to software pirates and hackers. Time to dig a little more into that.
Your standard hacker. Breakpoint generated this by taking a grainy copy of my Slack profile drawing and feeding it into Midjourney with a bunch of "8-bit" keywords.
Software piracy is all about distribution channels. Digital goods can be reproduced infinitely at will, and the only way to exert control over that is by locking down the distribution channel -- something pirates hate. There has always been a tension between corporations who insist on a moral right to charge whatever they want for software and distribute it with very limited "use rights", and pirates who want to see software get into the hands of everyone, no matter how poor.
(If your heart is too hardened to this whole Robin Hood scenario and you side with the corporations, you might want to consider how the same war is being fought with pharmaceutical companies who, having invented a drug that could save millions, price it as high as they possibly can, soaking insurance companies and taxpayers and leaving thousands of poor people to die. Is it their right? If you're a straight-up capitalist, yes. Is it still their right if they developed their drug using government grant money? Is it still their right if they use the income to lobby for tax loopholes and deregulation? Capitalism without roadblocks is a direct road to oligarchy, and software doesn't cure disease, but access to it can translate directly to power.)
I'm describing all this here to put a little positive spin on a group that has shifted over the years to be seen as entirely negative. These days, hackers and pirates are associated with scams and political unrest and destruction. Right or wrong, the first - and only - question people ask when they consider generative art tools being used by the unwashed masses is, "How much damage are they going to do? How afraid should I be?"
A Soviet computer enthusiast, conjured by Breakpoint.
The assumption here is that if the tools are in the hands of a large capitalist organization, they will respond reasonably to regulation, and are therefore "safe", whereas J Random Hacker could use the tools for anything, which is inherently "unsafe". That assumption rests on another one: That market forces always push towards the betterment of civilization, so it's in our interest to keep the market strong -- and profits high. Or put another way, if a company has found a way to derive profit from a process they have assembled, it is their absolute right to make that profit, regardless of the side effects of their process.
In this era it's very popular for corporations to wrap themselves in banners of progressive politics, claiming to be champions for free speech, for the poor, for equal rights, for criminal justice reform -- you name it. But the bottom line is, if you threaten their revenue stream, or threaten the image they've cultivated that makes them appealing to consumers, you will be quite casually plowed under. If not by the deliberate action of an employee, then by the machinations of the system, acting as it was designed to, without conscious thought. And if crushing you does not have consequences, no one will care. Profits are being made.
Let's pull these abstract ideas down to the ground: Say a company trains up a progressive art generation tool, and feeds it indiscriminately with art. They claim that they aren't violating anyone's rights because none of the original artwork they feed in remains after processing is done, so they are essentially "learning a style". Then they wire it up to the internet and sell access to people who need illustrations made with a short turnaround time.
"Gandalf Riding a Corgi Into Battle", whipped up by Breakpoint in only a few minutes.
You can imagine how a service like this would change the value of commissioned art: It would chop the floor right out from under it, while at the same time making certain kinds of art that were previously rare because of their technical difficulty much more common. For example, realistic-looking prototypes:
"Hellraiser Puzzle Box, My Little Pony Edition," whipped up by Breakpoint in about five minutes of iterating.
That's a pretty good service and revenue stream for a company, especially since they don't have to hire any actual artists. At most, they just need art critics, who know how to label things so the various styles and contents of the incoming art can be reflected in the text prompts on the other end. As long as they can stay ahead of the CPU demands for re-training their system to keep it current, they've got a profitable business. And it's built directly on the backs of artists, who in all likelihood will get paid jack squat for providing the essential material.
You can imagine other companies joining the market, racing to the bottom in price, and pushing the edges to see what else can be reduced to an AI-learned "style". For example, the fashion industry:
Arnold Schwarzenegger crossed with Stephen Hawking, from Breakpoint's spooky lab.
The concept here is pretty fun, but check out how the AI has also crossed Hawking's suit with Arnold's bare chest, resulting in a kind of sleeveless dress-shirt and sweater-vest combo that could easily be the foundation for a costume in a film or play. That was a happy accident. What could you produce with a tool specifically trained to remix clothing styles, ancient and modern? You'd still need a craftsman to produce it, but the inspiration would be much easier.
The Wienermobile Mark II.
This took a fair bit of tinkering by Breakpoint. He did a five-way merge of images he'd assembled with other settings, generated in the depths of Midjourney. It still amounted to about half an hour of work, though, which is less than it would take to gin something up in a paint program, and nothing compared to the time it would take to actually build a working model and see how it looks in a driveway.
So if companies had their way, they would perpetually coast on top of the output of small-time artists, vacuuming up their work as soon as it's produced, and then making digital crossbreeds on the cheap for anyone with cash. The only real conflict would be over whether one corporation can pirate the style of another. If you wind this forward five or ten years it's easy to picture a marketplace where art production houses hungrily cannibalize each other by training their AIs with each others' output, and the one that wins is the one that can throw the most CPU cycles at the problem and consolidate the most "style." People who come up with fully original styles will be operating in a sort of bespoke market that's subordinate to the great style vacuum AI enterprises that serve everyone else.
I don't know if this is actually a better future. A world saturated in art cranked out by machines doesn't sound like a more interesting place to live. We already live in a world where digital preservation can make a piece of art produced one time, in one place, last an eternity. Recorded music has been that way for decades now. You're not going to hire a contemporary band to play you some toe-tapping tunes in your living room, but you're going to put some jazz on the stereo because it's cheap and easy. The composer, arranger, and all the musicians who put breath and movement into the instruments, as well as the instruments, could be long dead and recycled into tin cans and fertilizer, but the music will be there, filling your ears with the style of a particular era. Leaving aside the debate over whether this situation makes contemporary music more or less vibrant or profitable, it certainly creates a dragging effect on culture itself, giving the past a greatly magnified influence over the present.
So, now that generative AI is on the scene, is a large part of the visual arts going to get permanently anchored in the styles that are available right now, during its ascendance, because independent innovation is going to stop being cost effective from here on out?
I think that's the way things would go for sure, if this tool remained firmly in the hands of for-profit companies with deep pockets and shareholders to please. But now let's involve those nefarious hackers and pirates and other creative people.
If it's okay for corporations to be style vampires and plunder the art world, then it's got to be okay for individuals to do the same, and anything that a corporation would not consider profitable to produce, or would consider damaging to its brand, could still be produced by an individual. An open-source toolset that could make endless remixes of artistic works old and new, will doubtless lead to new forms of art on the small scale that the corporate world couldn't be bothered with.
What the heck is going on in this scene?
Imagine using generative art to set the scene in a table-top game. You could feed in specific art, from your own favorite games and movies, along with photos you've taken, and throw together illustrations scene-by-scene, and draw narrative inspiration from what pops up randomly in the output. I mean, look at the scene above. I personally find it compelling because it's a mashup of 8-bit gaming art, noir-inspired lighting, and creepy horror themes. To me it just begs for an imaginative backstory that I could turn into an entire adventure.
But, your own tastes might vary, and you might find the same thing confusing and dull, which is entirely the point. You might be more inspired by this weird mashup instead:
Boba enjoying a beer in barbados.
This example re-raises the specter of copyright, but not in a way that isn't already covered by existing law. Just about everyone in the dang modern world can guess what popular character this is supposed to be, but if you're telling your own story to your friends, fair use law makes that irrelevant, whether the likeness was produced by hand with a physical brush or churned out by a computer in a few moments.
A wookie wearing wicker in a western.
"Fan fiction" has been around a very long time and this will enhance it. No reason you can't mess with stuff trademarked by congress either, if you're not engaging in trade:
The most irresponsible version of Smokey Bear.
(As an aside, Breakpoint was constantly running up against the limitations of the Midjourney toolset. It took several dozen outtakes to arrive at this semi-coherent version, and he was never able to get what he really wanted, which was Smokey holding a flamethrower.)
So it's not for sale -- so what? The amazing thing is, since the tool is just a download away, the value being conveyed is almost entirely in the idea, and not the execution. The execution is so cheap it's practically irrelevant. It's the Moog synthesizer of the visual arts: You feed a little electricity into it, and it can reproduce the effect of almost any instrument in any environment, assuming it has been trained with samples to do so.
And like an ordinary instrument in the hands of a jazz musician, a little improvisational flair goes a long way.
What an old computer magazine ad looks like when you're possessed by demons.
Imagine a gallery where an artist presents their works on digital frames, each generated with a custom trained AI. At the entrance to the gallery you are asked to describe a dream you had recently, in just one sentence. Then you're invited to walk through the gallery, and as people approach each frame, their dreams are combined in a re-generated version of the art. The evolution of the art is recorded each day and turned into a slowly evolving video, which is then packaged into another frame that people can buy and hang in their home. A time capsule of the dreams of the people who saw the art with you, on a certain time and day. An interesting notion, and up until now, utterly impossible.
Let's push a little further out into the realm of stuff that people might do, but corporations would be sued for:
You can imagine this being liberated from some 1930's CIA archive.
The internet has a long history of being used to prank people. Facebook alone is littered with a hundred thousand bits of Photoshopped junk, trailing long tails of "likes" between credulous n00bs, all of it fabricated. One might think that generative tools will bring a tsunami of junk, and skeptical and rational thought will be permanently drowned, bringing about the end of free society and the destruction of free speech as people flail to purge it from the world. A tsunami like that is definitely coming, but try winding forward a bit more, and think about where that goes.
None of this exists.
It took Breakpoint about half an hour of messing around with keywords and blended images to generate this. It looks really, really creepy, and that was exactly what he was aiming for: All the visual signifiers of a photo taken in some monstrous lab, from the depths of a conspiracy theorist's nightmares. This image could easily find life on Facebook, hopping around various newsfeeds, inspiring shock and confusion as it goes.
Half an hour to generate this image, and this is a tool that is still in active development, among others. In fact as of right now, this image is obsolete because a new version of Midjourney is available that does a much better job with hands.
How long will it take for everyone, or at least most people, to realize that they just can't trust any random image off the internet, and instead they really need to pay attention to what entities - corporate, private, political - are backing the information? After several generations of people not thinking or not even caring where they get news, or what they consider truth in their social and civic lives, because they're too enchanted by the idea that the information is "free" and they instinctively know truth from fiction, how are they going to adapt when they realize that anyone, anywhere, can make a realistic image that is an utter lie in about the time it takes to make and eat a sandwich?
Putin wrestling a bear in Red Square. Which definitely never did happen, and never will.
For decades now, society has been buying into the promise of the digital age as it's been marketed by the largest corporate players in it: It's a family-friendly information resource, full of useful and accurate facts and well-categorized entertainment, and a clear conduit between your eyeballs and the lenses of cameras, whether a thousand miles away or just across the room. That's always been a lie, especially in the realm of social media, where your ostensibly clear view of the world and your friends is actually chopped into a fine atomic powder and then reconstructed with various levels of censorship and promotion applied, before it's fed back to you and everyone you know... But that lie has mostly gone unnoticed, because many employees have been working very hard to make the deception feel irrelevant, or even necessary. After all, the internet is a conduit from anyone to anyone, and what is a social media company for, if not to act as a filter to block out all the people you don't like, who might upset you or cause you harm? Thus the notion of the public square has been transmuted into a micromanaged software driven apparatus where free speech is supposedly the effect, but safety from discomfort is the actual overriding goal -- because if this "public square" makes a customer uncomfortable, they will disengage.
So what happens when this "public square" gets so overwhelmed with false information of all kinds that even highly-trained machines cannot keep up with filtering it out, and ordinary people mucking around with generative art get plowed under en-masse by censors but have such fun with it that they keep coming back?
Putin up to his neck in Ukrainian soil. Which is definitely happening, even though this image is fake.
Two things. One, the filtering becomes much more overt, exposing the whole business model of social media for the cracked mirror it is. And two, ordinary people begin to actively seek out some mechanism that clearly delineates the line between communications from flesh-and-blood people they know, and robots, and the line between information produced by entities with an interest in the truth, and entities who profit from blurring that line.
A few specifics to this change, off the top of my head:
Publishers and editors will make bold declarations that all their content comes from real humans -- even the ones that freely publish fake images and use machines to ghostwrite their copy. The outcome of this deliberate confusion will be a digital certificate system similar to the ones used to verify the provenance of entire websites, but instead deployed on a per-item basis by publishers as a mark of their integrity. I.e. if it's not digitally signed by The New Yorker, it's not up to their standard.
Midjourney has clearly been trained on a lot of New Yorker covers. The words and the font are pretty well imprinted, even when the rest of the cover is nonsense.
This alone won't really change the status quo, but with a much more intense emphasis on the value of certification, there will also be more emphasis on presenting information in a context that conveys and guarantees the certification. People will realize they
need to know the difference between what's real and what's fake, and providers of real information will be in a better position to dictate terms about how their content is re-used. This will give them leverage against search engines and social media giants, both of which have a tendency to scrape the content out of other organizations and repackage it, hollowing out every other revenue stream but their own.
The tools to read the digital "chain of trust" in a document will be much more flexibly integrated with media consumption applications. Currently they are used almost entirely for copyright enforcement. In the future they will be used as part of the filtering and/or censorship settings of a social media app. Eventually all anonymized data - data without signed encryption from a trusted source - will be treated as hazardous and silently ignored by default in consumer communications software. Even if it's just a photo of your cat; if it's not digitally certified, it doesn't exist.
In the short term, Canon and Minolta will introduce cameras for journalists that digitally sign every raw-format image they take internally, establishing the provenance of that original for later verification with publishers. In the medium term, this sort of signing will be introduced in software to every smartphone. In the long term, the digital signature applied by the devices you own will be derived from a digital fingerprint that is part of your
government issued identification. (By the way, you really should see the movie "Idiocracy". It gets more relevant every year!)
Are you used to the safety of anonymity in crowds? That safety will soon be deemed a public nuisance.
So, this is my prediction for where we're headed. The proliferation of fake machine-generated tripe online will not end society but it will definitely transform it. Meanwhile pirates and hackers will use generative art to make statements, be creative, fight corporate monopoly, and eventually renegotiate the artistic process. I sincerely hope that some room for individual liberty remains at that point, though I do look forward to the renewed interest people will show in using physical presence to verify trust.
Hopefully we'll still have decent job prospects after all this...
![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
[00:54:47.97] "I'm in a little bit of trouble. Oh yeah." (gibberish) (slurping) "Why isn't he dying? What is he eating in my ear?"
Search engines used to take in a question and then direct the user to some external data source most relevant to the answer.
People learning about evolution sometimes ask, "Why aren't animals immortal?" The answer is, the world keeps changing, and life needs to create new bodies to deal with it. What we really want when we ask for immortality is one constantly renewing body, running all the amazing interconnected systems that we're used to, and that convince us we are alive from one day to the next, without interruption. ... Well, except for sleep, which is a weird exception we have decided to embrace, since going without sleep really sucks.
As a sort of last hurrah before my work schedule gets intense again, I played my way through 
Then I realized, the only way to frame an answer is by considering the act of programming within the act of communicating via language, which is something humans have been doing in very complex forms since there were humans, and in less complex forms before that.
This is not "intelligence". Whatever intelligence is active in this system is the intelligence that a human is bringing to the table as they manipulate a prompt in a generative tool to get some result, or change the way it classifies subsequent information.
For quite a while I've been looking for some nice way to get a complete backup of my Dreamwidth content onto my local machine. And I gotta wonder... Is this not a very popular thing? There are a lot of users on here, posting a lot of cool and unique content. Wouldn't they want to have a copy, just in case something goes terribly wrong?
So. Arrrr! Hello me swarthy crew, and welcome to the rant. Before we get fully under sail, I need to pause at the edge of the harbor and ask all hands a question:
Meanwhile, software developers are charging you a monthly fee just to have their software installed on that device. Regardless of whether you even use it. And they are getting away with this. You can call that "what the market will bear", yes. Go ahead. But I call that a bloated piece-of-garbage market that is already too cozy for software makers.
Fellow dork: This is the age we live in. The age of the nerd flock.
