Jason Allen won a prize for this digital ‘painting,’ which I am flagrantly ripping off and posting here without paying any licensing fees.
He was triumphant and a bit cocky about his win.
Much consternation spread throughout the artistic community two years ago when Jason M. Allen, an executive at a tabletop gaming startup, submitted an AI-generated “painting” to a Colorado digital art competition and won. Critics claimed that Allen had cheated, but the prize winner didn’t have much sympathy for his detractors: “I’m not going to apologize for it,” Allen said. “I won, and I didn’t break any rules.” He also didn’t seem to care much for the complaint that AI companies like Midjourney—the one he used to create his “painting”— were poised to destroy the art market. “This isn’t going to stop,” Allen told the New York Times. “Art is dead, dude. It’s over. A.I. won. Humans lost.”
Except that now he is dismayed to discover that he isn’t getting the rewards he thinks he deserves.
Now, in an ironic twist, Allen is upset that his work—which was created via a platform that’s been accused of ripping off countless copyrighted works—cannot, itself, be copyrighted, and is thus getting ripped off. In March of last year, the U.S. Copyright Office ruled that work derived from AI platforms “contained no human authorship” and therefore could not be extended copyright protections. Allen has been trying, since late 2022, to register his painting as a copyrighted work.
Last week, Allen filed an appeal in federal court in Colorado, arguing that the U.S. Copyright Office was wrong to deny copyright registration to his work, dubbed “Théâtre D’opéra Spatial.” Allen’s primary concern is that he’s not making enough money from the work. “I have experienced price erosion in the sense that there is a perceived lower value of my work, which has impacted my ability to charge industry-standard licensing fees,” he told Colorado Public Radio.
It’s so unfair. He worked so haaaaard on his picture, as if people should be compensated for how much effort they put into something.
Allen’s lawyer, recently claimed that Allen had worked hard on his digital illustration. “In our case, Jason had an extensive dialogue with the AI tool, Midjourney, to create his work, and we listed him as the author,” Pester said.
Sorry, dude. It’s over. Capitalism won. Humans lost.
As long as they keep their digital paws away from the book covers by Boris Vallejo, Julie Strain or Richard Corben…
(yes, I am a nerd)
Art is pretty much unkillable. Its an abstract.
Do they use “prompts” to generate AI artwork? If so, perhaps he could copyright his prompts. I gather that crafting generative AI prompts is an area of expertise in the field.
Now where have we seen this before?
Steve Jobs 1996: “Good artists copy, great artists steal. And we have always been shameless about stealing great ideas.”
Steve Jobs 2010: “We think competition is healthy, but competitors should create their own original technology, not steal ours.”
@2: art is indeed immortal, but artists sure ain’t. They do need to eat at least, and all this AI kerfuffle is going to make their paid work chances even smaller (the stereotype of the starving artist is often all too true).
I am putting this here again for a very good perspective:
(An aside: google did not want to show “vimeo wizard of AI” at all and came up with unrelated crap, while duckduckgo found it at once even if it was placed only fourth or so in the list. Whaaaats going on with them search engines?)
Since the ai created the art, the ai should get the copyright, but, of course, only if it applies for one.
It’s not even a particularly good AI render! The walls are full of weird artifacts, the window/balcony thing is asymmetrical, the people have weird postures or look fuzzy or distorted. The lamp on the right is fused to a malformed chair or couch, and the ones on the left are just… weird abstract rectangles. This isn’t the kind of AI output that should confuse people, this is the kind that idiots mass upload to wallpaper sites.
@ ^ tallora : Also are they walking on what? Water? ice? Really badly laid carpet?
@5.outis : “@2: art is indeed immortal, but artists sure ain’t. They do need to eat at least, and all this AI kerfuffle is going to make their paid work chances even smaller… “
Truth. Yup.
#7 tallora
Speaking as one of the visual artists he was gloating over, I’m completely unimpressed. Even if this had been done by hand I wouldn’t have been impressed because it’s nothing more than visual gibberish. Yes, the computer did all the work, and the computer is also completely untrained in things like composition and technique.
Actual digital artists make choices (about sizes, details, colors), that put more work into their art than this guy did with some prompts.
outis @ #5 — I note that the video seems focused on “generative AI”, at least in the first 5 minutes. Many people, including Marc Zuckerberg and people around here, tend lump together all the different areas of AI. While there is certainly overlap between the different areas of AI/ML research and implementation, generative AI is the new-ish endeavor that’s getting the most public attention right now and perhaps the most troubling because the leap from an statistical analysis of a large body of content (LLM) and turning into a generated response to a specifically query is obviously fraught.
There’s an interview with Janelle Shane in Scientific American titled, “Please Don’t Ask AI If Something Is Poisonous” where she talks about her investigations into generative AI and some of the issues she’s run into with it.
@5: As Google owns YouTube, it might be a bit biased about the competition…
@10, robro
I think people, myself included, lump Generative AI in with all the other subdisciplines of AI/ML because we’re not really talking about the technology. We’re talking about the tedious hype cycle, where hucksters, grifters and fraudsters both criminal and corporate all abstract away the complexity of the underlying issues and just point to a pretty picture (with only a few horrifying hallucinations) or a coherent sentence (that’s only slightly wrong) and say “True machine intelligence on a human scale is just around the corner. Give me vast sums of money to either build it or fight it.”
For artists, this is a difficult time. Generative ‘art’ production is presently cheap and good enough to replace a lot of low-value art. Think book jackets for vantity-press novels for example. But this isn’t guaranteed, or even likely, to stay the case forever. The companies providing these AI art services are burning through vast amounts of investor cash and investor compute resources (for example, most of M$FT’s investment in openAI is in the form of cloud computing credit). Sooner or later the investor class is going to want to see a return on this money, and if they don’t, the generative AI faucets are going to start costing a lot more to use or running dry.
The evangelists of this technology are demanding absurd amounts of additional investment, and even demanding that new kinds of power generation must be made so that they can fulfill their dreams of the god-in-the-box. Each generation of the models requires a huge leap in training data and computation to train and to run, and the results of each generation appear to be entering diminishing returns territory.
Sooner or later the investor class is going to realize that they’ve reached the top of the S-curve, if they haven’t already. At that point the hype machine will go into overdrive. They will, after all, need to recoup their investment. And the way to do this is to convince some greater fool that the real profits are all still to come, that out of benevolence the investors are going to allow the plebeian masses to get in on the ground floor by buying some stock on the open market. Once enough of the bag has been handed over to to greater fools to hold onto, the whole edifice will be allowed to collapse. See also: gold rush towns in the Sierras and the Klondike. The investor class will already be moving over to whatever hype cycle comes next… from Web3 to Metaverse to AI to ???
At the end of it all, the technology will remain. But it’ll be expensive to run. That ‘artist’ with his ‘painting’ – how many queries exactly did it take to generate the image? What would that amount of GPU time have cost in an unsubsidized market? Would he be able to compete with an actual digital artist who can paint something like that on a drawing tablet in a couple days and not have it full of weird anomalies? Who can say.
Personally, I think that generative image AI will end up as tool mainly used by artists. In-painting unimportant areas with detail, for example, or generating a starting point that the artist iterates upon. It’s not going away, but I doubt it’ll be world-changing.
In short: copyright for me, but not for thee. The pattern appears to be the same every time.
I just created four AI images to put on a birthday card with a ten word prompt. The recipient is my niece who, as it happens, is a digital artist — one who uses math and computers to create art installations, putting her masters in Maths and Theoretical Physics to good use…
I did it as a joke, but once I stopped trying to generate images in the “realistic” style with its horrors of misplaced limbs, fingers, and eyeballs, the results weren’t half bad for use in a casual birthday card, and perhaps easier than scouring Google images for a photo to swipe.
Seems to me that the days of clipart websites are numbered.
In answer to one of the items above:
The prompts to the generating engine are probably not, themselves, copyrightable. Copyright doesn’t protect original ideas; it protects original expression (which can include original arrangement, selection, and analysis of facts — thus, Dr King’s “I Have a Dream” speech was copyrightable, although not one “idea” in it was novel, leaving the real question to be who could copyright it, but we’re getting away from the issue). The shorter an expression, the less likely that it can be protected by copyright. Although Copyright Office regulations are not definitive on what is “copyrightable,” they explicitly refuse registration to short phrases — which is precisely what most generating-engine prompts are. This is consistent with the last century and a half of both legislation and judicial decisions about “what can be protected by copyright?” throughout the world, so:
I’m fairly confident that prompts to a generating engine are not protected by copyright, except in extreme instances in which the prompts were long, involved, and displayed creative expression independent of the output.
I’m not buying the notion that the dreck generated by AI could be called art. I’ve seen better art made by elephants and chimpanzees. This is merely a poorly rendered digital rip off of the moongate in game of thrones and medieval art.
Much like the horror that is the Rings of Power, the cgi effects ruin the show, even without the terrible writing and fan-fiction plot. Tolkien would hate it too.
@ #5 robro and #15 Jaws
Right now on display at the Los Angeles County Museum of Art is a generative piece of abstract art in which the instructions on how to create it are what is copyrighted and owned by LACMA.
Typically they choose a wall, and paint it on.
A good, non-committal comment to make when viewing bullshit “art:”
“Boy, that looks like it took a lot of work!”
@17: A copyright notice does not make a valid copyright claim. Museums — and, in particular, museums tightly focused on “fine art” — are notorious for overclaiming what rights they have/might have. For example, there’s an entire line of cases well-known to copyright nerds like me that denies the museum copyright ownership perforce in photographs taken of works on display (even works that are themselves in copyright and for which copyright has been properly transferred to the museum). That doesn’t stop museums from making that claim; take a look at the back of the admission ticket for A Certain NYC Museum…
Not having seen those actual instructions, it’s hard to say as to the particular LACMA example, but this in principle is just another iteration of the “list of ingredients versus constricted-by-convention instructions versus commentary-on-the-recipe’s-origin” brouhaha that has been definitively decided against copyrightability of the first two… for decades. One might try to compare it to “copyrightability of computer source code,” but then we’d be getting into law-journal-article territory and certain decisions (one by then-Judge Gorsuch) firmly against copyrightability of the instructional aspects of the source code.
So I’m not all that worried about what LACMA could do if I input those instructions into the same generating engine. I’d just have to spend an hour responding nastily to the C&D letter their outside lawyers would send…
One final note:
There’s no such thing as any “industry-standard licensing fee[]” asserted by Mr Allen. There are customary starting points for negotiation, but wide variance even there. General rule of thumb: If deviation from the “standard” cannot be prohibited or controlled, it’s not a “standard” in the first place.
Bluntly, he’s in the same boat as someone disappointed because the Caravaggio they’ve been trying to sell at an auction house gets devalued because someone asserts it’s a forgery.
(Just in case you can’t tell, I don’t have a lot of respect or sympathy for the special-snowflake aspects of the so-called “fine arts.” Yes, artists do want and need to get paid… but so long as the means of them getting paid is “proceeds from individual sales and licenses,” they’re not special just because they’re doing “fine art” instead of making any other original creation.)
Seems to me that the days of clipart websites are numbered.
Ah. clip-art…I remember when we had to walk through three feet of snow to buy clip-art on a CD in a computer store! With no snow-shoes ‘cuz we didn’t have any of them newfangled weather apps to tell us it was snowing! And we liked it that way!
“I have experienced price erosion in the sense that there is a perceived lower value of my work, which has impacted my ability to charge industry-standard licensing fees,” he told Colorado Public Radio.
Since when was “ability to charge industry-standard licensing fees” a legal right? And what if I haggle a merchant down on the price of something he’s trying to sell me? Could he then sue me for “price erosion?” Or could he sue someone else who might have advised me that the product wasn’t worth his asking price? I see a major conflict between free speech and merchants’ ability to silence critics of their products.
The prompts to the generating engine are probably not, themselves, copyrightable.
Those are nothing but specifications, demands or wishes. When have those EVER been copyrightable? If I write a sentence like “Draw me a picture, in the Pre-Raphaelite style, of a dark-haired girl in a long white dress sitting on a log in a forest and crying”, would that be copyrightable? If I did copyright it, would that mean anyone who drew or painted a picture following those specifications owes me royalties? Or would I have to prove that they had both read my sentence beforehand and consciously followed it?
Perhaps in visual media but in, say, writing a story, the prompts can be quite specific. For instance, I’ve used AI to write stories but I used characters, including their names, I had created myself. Sometimes I’ve used fictional settings I created myself and had the AI craft a story around it. I would argue a work like that should be copyrightable.
I like Thor, the goblin king’s, take on AI art. You can charge for AI art, but only after you reimbursed the artists whose work was used in training the AI.
Also the moment you start making bank with AI art is the moment AI tech companies will start vacuuming your money out of your wallet.
gijoel,
I don’t think you get how LLMs work.
They are not copying stuff, they are generating stuff.
As for the dataset at hand:
https://www.artnews.com/art-news/news/midjourney-ai-artists-database-1234691955/
—
The funniest aspect is that Jason’s claim to artistic merit is being able to type in a prompt, and then fiddle with it.
I concede it’s true there is some merit to his claim; basically, about as much as solving a chess problem by asking a chess engine to do it. That’s not a great amount.
@12 “For artists, this is a difficult time. Generative ‘art’ production is presently cheap and good enough to replace a lot of low-value art.”
Well.. A) its only cheap in the sense that no one is looking at the costs of running it on online servers, then.. its expensive as hell. B) If it was good enough there would not already be companies out there that are promoting their stuff as AI made, while hiring artists to “fix it” behind the scenes, at low pay, and hoping no one notices. Otherwise.. eh, maybe. In the long run its likely to just become a tool for iterative design, with the final product being human anyway. There is a guy on Youtube who described his new process of producing art, compared to the old one. The new one is, “Take a picture(s) of my wife, who I use for all my art, come up with an idea for a character I want her to be, feed that into the AI, then pick from various outcomes which one I think looks best, hand edit it to fix mistakes made by the AI, then feed this back in with a few more ideas, like adding more machine like parts to the character, etc., rinse and repeat, until I get close to the final result, then clean that up, by hand.”, or, “Have an idea, draw dozens of variations of it, throw most of them out, then either spend weeks trying to get the final result right, or get so frustrated with the process that it ends up in a desk drawer for the foreseeable future.”
For him, its a tool. For the idiots just feeding prompts into the thing, and having it spit out stuff, which they then just feed back in, maybe…, to tweak more, while never doing any work themselves, its just twisted garbage. And.. its extremely unlikely, no matter how much training or tweaking done on the engine, that it can ever get that much better. Heck, the “major leap” we saw not long ago was a result of them releasing two versions, back to back, very close together, because they had been working on both for months, and decided to throw something out, before it was really ready, to spark interest, before throwing out a new version that was, ironically, still the work of months of messing about to get it working. And there is no evidence we are going to see some massive leap again (or that a real one, such as would be necessary for it to comprehend what the heck it is even trying to create from the prompts, and thus avoid mistakes doing so, is even possible to achieve with LLMs).
I have prompted a large number of AI images, some of which I have liked enough to post on social media. But I have never had the temerity to call myself an “artist” for it and have always found that hubris laughable and pathetic. Actual artists have long had their work diminished by others, “Why do you charge this much? It’s a hobby, not a job. You should be giving this away,” but this desire to claim the title of “artist” shows just how much of that is the envy of people with real skill and talent who have put in the work to develop that skill and talent.
Anyway, this will all be a moot point once Disney wants to start copyrighting generative art and lobbies to get the laws changed.
@kagehi, 26
That’s what I meant – it’s presently cheap, because everyone in the investor class knows that to have any hope in hell of monetizing this thing they have to get people using it and integrating it into their lives and workflows in such a way that when they start (gradually, softly, slowly) raising the prices that people will keep using it.
They need people to become dependent on it, and each of the companies involved would like to have a monopoly on it, before they start to turn the screws and profit-take. See also: Uber. At the moment, it’s being heavily subsidized. Every GPT-created troll comment, every low-effort AI shrimp-jesus used to befuddle boomers on facebook, every shitty deepfake. Unless they’re created with self-hosted AI software, all of that glurge is being churned out courtesy of the investor classes, in their desperate hope to be ‘in at the ground floor’ of the next iPhone.
Your example of the artist using AI, with many many manual steps, retouchings, edits – that’s probably the future of generative image AI. Generative video, being essentially uneditable without vast resources is just going to wither to almost nothing (it might continue to exist in VFX houses…. though it doesn’t sound too promising).
I think we’re near the peak of the AI hype cycle now, if we haven’t gone past it. OpenAI has managed to raise $6.6B in this latest funding round, on a laughable valuation of $157B – for a company that not only has never turned a profit, but whose ‘plan’ to become profitable requires entire new kinds of energy generation and speculative advances in GPU technology to be possible. David Gerard on his excellent site Pivot-to-AI has covered this in all its absurd glory. Sure, some of those investors are professional money-losers like SoftBank, but others are going to want to see something in return for all that money they’re setting on fire.
And that’s why I said above that they’ll start an overdrive in hyping right around the time they start offloading their shares onto the gullible public. Insiders extracting money from Retail investors is how all the previous cycles eneded, I see no reason to believe that this one will go any differently.
@28
Yeah, OK, I suppose this is the new “early internet”. Every idiot on the planet is hyping it up, almost none of them understand it, or what it will be actually useful for in the long run, and 90% of all the nonsense businesses they promote through it will crash within the next 4-8 years. Then the only question is, “Is it creating a big enough boom that we will have another Reagan era style ‘prosperity'”? – i.e., taking credit for a huge boost in the economy, from investment in something that melts down over a single administration, but then some idiots keep insisting the person who was coincidentally in charge through it got us there via “genius economic policy”, which is actually both dumb luck, and doesn’t actually work.