Artofficial Inktelligence III: Copy; right? - A human-generated story in 3 chapters.

After I’d been on BBC Leicester to talk about AI, I arranged to visit Andrea Soltoggio, Associate Professor in Artificial Intelligence at Loughborough University.

Although I probably knew a fair bit about AI compared to Maureen, busy shopping in the High Street when she was vox-popped by the Beeb (although not necessarily: Mo could be knocking out thousands of Made-By-MJ Adonises every night after work Because She Can Now) I realised I needed to increase my knowledge all-round; especially in the area of copyright, a wiring-loom-style nightmare of confusion that’s currently the subject of myriad studies and manifestos.

©Maureen. Except, not, because this dude can’t be copyrighted. Maybe he needs a girlfriend, though?

Andreja had been on the radio just before me, so I went over to the university and talked for an hour and a half over decent coffee.

I’ve advised on copyright and IP, and obviously deal with it regularly for work, but what I knew about AI and copyright amounted to this:

• First, copyright is automatic at the point of creation — in this country (the UK, from where I’m typing) you don’t have to ‘do’ anything ‘to copyright’ an artistic, photographic or literary creation.
• Copyright is enshrined in UK law, poignantly enough, as a human right by the Berne Convention of 1887, still applicable in law, which states:
• “All (artistic and literary works) except photographic and cinematographic which have different terms) shall be protected for at least 50 years after the author’s death, but parties are free to provide longer terms if they choose.”
• However there are many countries who are not signatories to the Berne Convention.
• The Copyright Act states that the author of an original work owns the copyright to that work. At the time the law was enacted, the purpose of copyright was to act as a financial and intellectual stimulus: 
  “The Act exists to incentivise original authentic effort. The person’s contribution is protected.”
• This government page is a nicely readable, brief guide to whether and how your copyright is protected abroad, but since geography is irrelevent to AI-generated art, this is only so comforting.

One of the largest swells of research and discussion has been to work out whether copyright can be claimed by someone making a picture using AI , and conversely, whether the copyright of an artist whose image has been ‘scraped’ from the www has been breached, which would normally automatically trigger a conversation about financial remuneration. This might seem obvious now, but it wasn’t at the creeping outset of the current rabid AI use, when it was all ‘look, I put my face into the Lensa app’ — you could never own those beautiful, flattering but heavily stylised portraits of you, based robustly on existing artist’s styles.

So platforms have been making up their own rules, since there’s no umbrella policy to cover things. Midjourney’s looks like this:

• people on their PAID MJ plan theoretically have copyright granted in anything they make, since it assumed the fact that they’re paying means they have commercial interests or incentives in using the programme; at the same time, an unlimited licence to use the image is granted back to the platform itself, meaning ‘it’ can do anything with it:

• Anyone else is free to put in exactly the same prompts used by someone else —you’d get a different output anyway, because that’s what ‘generative’ means.
• people on the FREE MJ plan (who it is assumed are generating experimentally or just ‘having fun’) do NOT get granted copyright:

Unusually, the UK is one of only a handful of nations to offer copyright for works generated solely by a computer, but it deems ‘the author’ to be “the person by whom the arrangements necessary for the creation of the work are undertaken.” Interpret that as you will.

In the US, there is currently no copyright protection for works generated solely by a machine.

It’s the chain of creation of an AI artwork that makes it so difficult to determine where the work ‘originates’.

AI-produced works of art start with a programmer (or more often, a large team of programmers) who have written an algorithm. That person may — or may not — provide some input data. Another person - an artist, or group of people - might do this. The output is then generated by AI independently.

So if an AI-generated artwork isn’t copyright-protected by law, anyone can freely make copies of, distribute it, use for commercial purposes or sell the work to anyone else; it can also be remixed, chopped, imitated, and so on.

…which sounds exactly like what can and does happen to copyright-protected work, just without the financial or legal consequences.

Which makes me wonder; the rather cruelly-named ‘prompt monkeys’ describing themselves as artists and selling their output online — don’t they want some ownership over what they’ve made? If Maureen’s keeping herself in fancy earrings through her Etsy shop selling those meaty He-Men, would she be mad if Dolly in Pennsylvania started ‘borrowing‘ her screen-shotted He-men, but to fund her Louis Vuitton habit instead?

Possibly not. Maybe that’s going to be the big pivot — that all art becomes universally, omnipotently open-source; no-one’s, and everyone’s property. There’s something in that I like the notion of, in an incredibly idealistic woo-woo kind of way — but then, money. Paying the bills, buying food, you know; the things we do to stay alive. If we already lived in a world where the Universal Basic Income was a well-established fact of life, and capitalism wasn’t the system, something like that could be beautiful.

At least one Midjourney user has it all figured out:

“Watching Artists get mad 😡 at ai wondering why? I mean as an artist I don’t understand copy rights [sic] because I’m constantly wondering about influences and I think these artists concerned with ai copying their art is a compliment because you know every artist is copying someone else’s work to inspire themselves regardless original ideas are very slim compared to imagination. As an artist I’ve never seen a non influenced art piece besides imaginary creatures and they’re [sic] not many left to create… never see any landscape artists bitchhing [sic] about the same mountain scenery…”

(I’ll wait a sec while you spot the really special bit.)

This is why there are big moves to work it all out, before the alleged robopocalypse — known in scientific terms as AI Singularity — hits. And my chat with Andrea very much started out as a list of questions about copyright, but ended up a broader conversation about the philosophy of art, its purpose, the human intention, and much more besides. He talked extensively about which artificial intelligence projects are already redundant — the self-driving car was an example—and which aren’t, such as medical and scientific problem-solving applications. There are things that are un-AI-able too — something it’s easy to overlook when there’s a constant stream of doomy prophesies flying at your head.

By the end of it, I felt reassured and freshly curious about AI and its place in our lives in a bigger-picture way. This was a man who works with the machinery on a daily basis - his life’s work — and if he wasn’t afraid, why should I be?

But there’s still the fact that copyright and ownership of intellectual property is tied up with money; it’s how revenue is generated for artists, through royalties, licensing, re-licensing, adding work to stock libraries and more. And since copyright is granted for 50 years after a creator's death, the artist is not the only human to benefit financially from their life’s work — providing wills are left and all in good order, children, spouses and grandchildren can, too. This notion of artwork being stolen by AI and melted into an enormo-bowl of cake mixture from which billions of new cupcakes emerge doesn’t simply put the much-maligned Tarquin The Painter’s nose out of joint — it could have a tangible, terrible impact on the futures of his dependants and descendants, too.

“Legislators will take forever to even fathom the tech, let alone build in any guardrails, so it takes things like litigation and public debate to move the needle on the ethical use of this technology.” — Jason Chatfield, president of the National Cartoonists Society.

Hollie Mengert is an illustrator for Disney who found that her work was being cloned as an AI experiment by a mechanical engineering student in Canada. The student downloaded 32 of her pieces and trained a machine learning model that could reproduce her style — it only took a handful of hours. As she told technologist Andy Baio, who reported on the case:

“For me, personally, it feels like someone’s taking work that I’ve done, you know, things that I’ve learned — I’ve been a working artist since I graduated art school in 2011 — and is using it to create art that that [sic] I didn’t consent to and didn’t give permission for.”

Hollie’s illustrations, left, used to train the model, and the trained Stable Diffusion output on the right.

“I see people on both sides of this extremely confident in their positions, but the reality is nobody knows,” said Baio in an interview with The Verge. “And anyone who says they know confidently how this will play out in court is wrong.”

For now, our own government has decided with the input of the Designers and Artists Copyright Society to scrap controversial plans granting artificial intelligence platforms an all-encompassing exception to copyright, asking instead to work with creative industry leaders to strike a balance between promoting innovation within AI, and supporting creativity in the UK:

“The debate followed months of objection by the copyright and creative industries warning of the detrimental impact the proposals could have on UK creators.”

The words ‘could have’, written in February, already sit uneasily with me having watched colleagues talk openly about reductions in their client work on social media; others new to the industry talk in bewildered terms about a tentative career they’ve just got off the ground that’s already impacted. Others at the very outset of their career, perhaps still studying or recently graduated, are openly questioning the point of pursuing such a career — careers that, as of 2020, were contributing almost £13million to the UK economy every hour.

— — — — — — — — — — — — — — — — — — — — — — — — — — — — — — —

I doubt this is the last time I’ll write about AI, though to be frank it’s been exhausting, keeping up, reading and archiving articles and continuing to walk the tightrope between my natural excitement and curiosity, and outright fear.

Though I didn't react the same way, I feel for the very, very established and much-loved Dave McKean, whose response to playing with AI for the first time was visceral:

“I spent a day on the floor of my studio in a foetal position…My immediate response was: Well, that’s my career over. Why would anyone pay me to do an album cover when anyone can type a few words into Midjourney and, in a couple of minutes, start downloading endless finished possibilities for free?”

I thought about ending this article by summarising the idea of AI itself as a playful, obliging but not very bright puppy that’s been let into the park too early, and everyone wants to play with it but as a result lot of people are getting bitten and shat on — but that sounds a bit naïve, and not quite right.

“There are a lot of people whose argument seems to be: ‘Well, the technology is here now, there’s no taking it back.’ I personally find that ridiculous because you’re just telling artists to give up without a fight.” — Sarah Anderson

Instead I’ll quote my partner Leigh:

Making AI Art vs Making Human Art is like masturbation vs actual sex with another person. It’s like PornHub; you go in and type what you want (‘big boobs’, ‘forties’) and there it is. You might get the same outcome, but none of the feels, the textures, the scents, emotions; you feel your way through.

Like a PornHub user, a prompter is always still sat safely tapping a keyboard. It has its place. 
But an artist is in a way more akin to a porn actor; willing to go the whole way, in person, for real, many times over, to get the result they want.

For now, I’m still refusing to think of AI (as it relates to art) as a monster that’s escaped from under the bed. I enjoy playing with it. I still think it’ll be useful to me and I watch in awe as people use the machinery to generate beautiful, beautiful things. The following pictures made me gasp, and almost brought a tear to my eyeballs — and if that’s not an example of a picture triggering a human emotion, howsoever created, what is?

Further reading:

https://www.mybff.com/discover/the-risks-benefits-and-ethics-of-artificial-intelligence

https://theaoi.com/campaigning/campaigns/consultations-blocks-2/artificial-intelligence/

https://theaoi.com/2022/12/09/artificial-intelligence-ai-aoi-survey-results/

Maybe this will interest you:
https://theaoi.com/event/aoi-webinar-ai-and-illustration/

PS: At the time of writing, AI is still struggling with type.

But the spectre of Banksy’s chimps is never far away.

Artofficial Inktelligence: A human-generated story in 3 chapters: 1/ Mad Journey.

This article’s been a long time in the making.

Not being able to ‘get to it’ has hassled me for ages but, as I finally get my assorted notes and thoughts assembled, I realise that quite unintentionally, the right time is now.

I’m writing about AI. Of course I am. Not only am I expected to have an opinion on it, as an artist (I do), but every man and his hamster knows about AI and moreover, also has an opinion on it. This is no bad thing; but you know that something’s ‘made it’ into the public consciousness when you’re asked to talk about it live on local radio; literally ‘the man in the street’ invited to share their thoughts on the matter, before switching to Sarah on the mic.

But more on that later.

In 2015 I began to play with Google’s DeepDream. Our mate Kev told us about it. You had to input a picture and wait for it to render — a long enough wait that you could go away and do something else, like have dinner and sleep, as opposed to Midjourney’s amphetamine-rush of fired-out images made in moments. In fact, facebook tells me I uploaded it on 30th July, and got the result on 2nd August:

So this was my first AI picture, and it’s a horrifying belter:

Kev, Sarah and Leigh, after 3 days Deepdreaming.

We laughed and pulled faces at it and then I forgot about it, and about AI for the most part other than following a few scientists on tumblr who were doing silly and very funny things with neural networks; getting them to write recipes, for example (generating the bewilderingly unappetising meal equivalent of what you see above) and poems. I would chortle at their painfully awful but nearly-there attempts, helplessly projecting human concepts like ‘trying really hard’ onto what I knew was not an earnest seven year old, but early machine learning.

A bit later I got to try out Adobe’s Neural Filters in a new PS update. I shoved a picture of my face into it and played with the sliders that deployed artificial intelligence to change expressions and age. The results were deliberately awful, because I knew I would never actually use them, so I went to extremes and pinged them off to family for Whatsapp LOLs. Regardless; the very gentlest of alarm bells was ringing. If I can change a face to this extent so easily, what’s to stop someone else changing a face for nefarious purposes?

Fast forward to 2022. Our mate Kev Again — featured previously after his mauling by DeepDream — mentioned an AI programme he’d been using to create pictures. He was having rather a lot of fun with it, and I had a go myself. Hm, I thought, yeah, it’s quite fun! (“Pin all the things”.) But I couldn’t really see how I might use it in my work, even if I’d wanted to, and besides, my contracts for client work are pretty much 85% a giant sign held aloft that says Thou Shalt Not Submit Any Work That Might Ever Have Had Anything To Do With Anyone Else Ever.

So, NightCafé Studio was also just a toy, then, and I forgot about that as well. Besides, at that very moment we were too busy getting excited about NFTs to think about artificial intelligence (more on THAT later, too).

And then; out of the blue I got an invite to beta test an app called DALL*E, which made artificially generated pictures. As you might be able to tell by now, I’m never one to back away from the thrill of a new piece of kit or technology, especially one so apparently BRAND new that it isn’t even out yet, so I seized the day. I asked it rather predictably at first to make a picture of ‘a night-time street scene in an 8bit art style’, because I was deep into my vaporwave playlists that week. And it did. Not bad, so long as you don’t want to print it at any sort of acceptable resolution or pass it off as your best work:

I did it again; this time, ‘shopping for pens and ink in a haunted pen shop in the style of Aubrey Beardsley’. Hm; it made a decent fist of that too. In fact, I liked them quite a lot. They had a certain charm.

After a great deal of messing about, I realised I should really have a look at whether it could reproduce anything that looked ‘like me’ — I don’t mean ‘a portrait’, though that’s happened since (it didn’t work very well). I wanted to know if the system could produce work that looked like mine. For the first time, I got the barest inkling — pun entirely inktentional - of the theory of a notion of an idea of a suspicion that this might, at some point, perhaps be a threat.

Let’s go in with possibly THE most widely-known of my pieces of work, I thought. Let’s ask it for a cover for To Kill A Mockingbird. I could see from the results exactly what it had been trained on — I could even name some of the illustrators — but nothing was usable beyond rough concept sketches or layouts, with useless lettering.

I then asked for ink drawings of Christmas iconography — I thought I’d try some old and very mainstream work that’s been out there a long time. And it made some alarmingly naturalistic looking stuff. Immediately I was put in mind of all the times I’d been ripped off in the very early days of doing hand-lettering, by people and companies who should know better but were seeing hand-drawn type for the first time (we’re talking late 90s and early 00s) and weren’t thinking that lettering was capable of being plagiarised as well as images.

…hmmm.

Despite the minor shudder, I kept on. I asked it to make words with ink. Totally illegible, though the energy was kind of there. Then logos. Then ink bottles. The cross hatching. Then illustrations for horror stories for kids…in black and white. Then landscapes in ink. And on it went.

It did them all, with its only stumbling block being that lettering was entirely unusable and hands were snigger-at-the-screen terrible. This was still only about a year ago, but even at that point a tiny but chilling voice was whispering Banksy’s ‘Laugh Now’ into my brain.

Over the course of the rest of the year I kept on playing. Kev Again was integrating AI into his image-making, and nicely so. Hands continued to be grim. But humans in my creative orbit began to make uneasy noises. I love tech, especially the sort that enriches my work and my process — whether the ancient Adana, my 80s Gocco printers or the latest Adobe feature, and I’ll be damned if I’ll be threatened by any of it; try it, learn it, and find out how I can deploy it. Another thing in the pencil case. But…but.

Never mind. On with the work. As well as playing with AI I was simultaneously immersing myself in the Web3 world, buying crypto, building an NFT collection and spending sometimes too much time in Discord because such things, just like Apple Pencils, Photoshop and magical things before them, are always just the next step in how we make, share and sell things. Nothing phased me — peak excitement was reached when minting a hotly-anticipated NFT after getting a whitelist spot. WAGMI! It was later stolen, but…yep, more on that later.

Around summertime I dove into Holly Herndon’s Spawning project. A lifelong lover of electronic sounds, I knew her records and was enthralled by her conjoined-twin Ai+Holly progeny. At the same time, she and her colleagues had developed Have I Been Trained?, a site which claims to allow artists to choose whether or not their images can be scraped for use by Stable Diffusion, MJ and OpenAi. I thought long and hard about what to click, yes or no. Did I want my images included in any large image training model? In what feels like a lifetime ago, I had a rather generalised feeling at the time that I didn’t want to be *excluded* from the world of AI, so clicked ‘Yes’. But that wasn’t actually what was being asked, and by the time I understood the specifics of the question some time later and went back to change it to a ‘No’, the site had inverted its system to one where, rather than choosing to opt in or out, you had to manually seek out any images of yours in the system, and ask for them to be excluded one by one (that’s still the system they’re using today).

I carried on. Towards the end of the year the world of commissioned work began to feel distinctly odd, with subtle but long-in-the-making shifts in modus operandi, their origins in the pandemic, finally manifesting in the way that work was coming in (awkwardly, unpredictably) and being managed (slowly and painfully). I began to see colleagues talk publicly about work being ‘weird’, and turnover being down. Against this landscape, my social media feeds began to fill up with people excitedly spewing out picture after picture made mainly in Midjourney, some calling themselves…artists. I, businesslike as ever, shut down the usual quarterly roundup of IP violations on Redbubble and Etsy, annoyed by the usual brazen help-yourself attitudes to artists’ work in the public domain — something which feels quaint now — and continued to stay abreast of the NFT world despite a sobering wallet theft, a gifted NFT stuck in cyberspace and a shady ‘buy it back’ deal with an anonymous stranger.

In the run up to Christmas I was busy with the annual seasonal promotion, the demands of the daily Inkvent project and assorted client jobs. I didn’t ‘play’ with AI again until the new year, using it to generate some atmospheric 80s-style pieces as refs to try and capture the vibe of some upcoming illustration work, and to experiment with layouts for a looming book project. We used MJ to make targets for my Dad’s shooting practice — zombies who may or may not look like certain politicians, all OK because of course they weren’t real people — and introduced my Mum to AI, whose first Midjourney piece was a beautiful Japan-inspired coastal landscape rendered in ‘ink’. She was fascinated, and loved the process — “I can put away my paints now’”, she said. “So can I”, I said, trying not to let inherited gallows humour extinguish her lovely anticipation.

I made a part-AI-part-Photoshop-part-Procreate leaving card for a mate, and used it to help make a silly birthday card for my sister. It was useful for generating some demographically accurate reference figures for an artwork I was making. And I did some more of those 80s experiments. I even started an Instagram page to capture the best ones. Kev Again continued to deploy AI subtly and successfully in his design work, and although I still hadn’t found where AI might ‘fit’ into my work, I still felt like it was a friend rather than a foe.

And then…and then. I joined the Midjourney facebook page, and at the same time, people calling themselves ‘professional prompters’ (to a great deal of derision and hilarity) and ‘AI artists’ began writing about how ‘real artists’ had somehow been gatekeeping, hiding secrets and lying about how long it took to create work. Look! many of them said; it IS easy after all! How dare those artists charge so much to do this!

I love this picture. She’s beautiful. I made her. Or did I? I didn’t. I asked a machine to make her, in quite a lot of detail. I worked with a machine. Or did the machine work with me?

Almost overnight, the world seemed flooded with articles about AI: the threats, the warnings, rabid defence of it, declarative statements against, over-confident opinions thrown in all directions. People fought to defend their use of AI, to explain why human artists didn’t deserve to keep working, and others outlined how the world could never be without human-made objects and pictures. My stateside agency had begun to talk about it in their newsreels, and my Medium feed was suddenly full of tediously repetitive articles loudly telling me how to generate five-figure sums with a ChatGPT side-hustle. How to generate pictures in Midjourney and flog them back to people via Adobe Stock. How to bypass concerns about copyright to make AI pictures to print and sell on Etsy…oh, yeah, Etsy. Back to that old chestnut.

I talked about AI at length with my agency’s boss, collected opinions and followed real-world artists creating beautiful work using AI as one of their tools, trying to keep healthy, future-positive role models in my line of sight. I remained determined to be neither for nor against AI, neither scared nor cultishly absorbed by it. And despite wanting to pontificate in the public domain many times, I held off writing about it, not wanting to add another voice to the clattering hubris of prompters calling themselves artists, and artists throwing shade on users; the world did not need another article wanting to sound knowledgable and finite.

So I didn’t, until a single tweet brought me to the BBC’s attention.

Chapter 2, in which I talk about my research with regard to AI + copyright + IP, follows.