COMMENT: AI in the arts can never do as good as job as a human, and is only likely to be viable for formulaic and uncreative activities. Unfortunately, there’s a lot of people don’t have a problem with any of those things.
Panic! The Terminators are coming! And the Cybermen! And the killer robot dogs from that Black Mirror episode. They’re going to start a nuclear apocalypse and hunt the survivors and go “Exterminate Exterminate” and put you into dystopian computer simulations … but even worse – they’ll take your jobs as artists! It’s even in the next Matrix movie. As the Sentinels gather round Neo’s broken and bleeding body, Agent Smith gloats: “Before I assimilate you, know this, Mr. Anderson: it was I who ruined your career as a screenwriter. I generated the spec script that was chosen instead of yours! I saw to it you abandoned your dreams to work in a call centre. Bwuahahahaha! Bwuahahahahaha! Bwuahahahahahahahahaha!” [Jarring chords. Lightning flashes in background.]
That, at least, was the reaction of half the world last year, whilst the other half were marvelling at a new golden age and/or seeing an opportunity to be the next Elon Musk. But reality is swiftly catching up with last year’s hype. There have been some remarkable technological advances: the fact that computers can now understand requests written in English instead of a programme language is impressive enough – the fact that it can create plausible images, text or music in response is incredible. But compared to what a human can do, it’s not that good. At its best, it looks good to people who haven’t wised up its predictability. At its worst, it’s waffle, plagiarism, or both.
Nevertheless, there are some areas where AI output, if not better than human output, is cheaper and easier. For reasons I will go into shortly, it’s been over-hyped, and the technological advances are no more or less dramatic that previous technological advances. Artists adapt and make use of them, and there’s no reason to believe jobs are under threat – not unless we we’re happy to remove all talent and creativity from the artistic process.
Unfortunately, that’s a pretty major caveat. I will explain why shortly. To start with, however, here’s why we shouldn’t need to worry about AI – in theory
Why AI ought to be no threat to artists:
I won’t tell a lie, AI is not going to be good news for everybody. As with all technological advancements, there will be winners and losers. But look at all the technological advances made so far. Few people wish they’d never happened.
There are two things you need to be aware of about AI. “Artificial Intelligence” is a catch-all term for a huge number of technologies. Prompts like ChatGPT and image generators get a lot of attention, but much of the time, AI is imply a marketing buzzword way of saying “We got better at automation”. The other thing you need to be aware of is that Artificial Intelligence is not creative. Or intelligent, for that matter. What AI is good at doing is rehashing content that has already been made, including creative content. And, to be fair, humans also do a lot of rehashing of creative content – it’s just that AI can do it faster.
I’ll start with the most controversial one at this moment: image generation. At the moment, there is an argument over ownership of images and the source material. A lot of AI art is analogous to tracing someone else’s work. Some organisations are getting strict over this. In video games Steam has made a rule (correctly, in my opinion) that any artwork that appears to be AI-image generated is assumed to be copyright infringement unless you can show it isn’t. This should safeguard against IP theft and also provided a respite against artists worried about AI taking their jobs. But the latter is only a temporary reprieve. Image generators such as Adobe Firefly work entirely with source material that is either public domain or owned by Adobe. You might have moral grounds to complain that somebody used freely-available images instead of paying you. You certainly have no legal grounds.
Here is my unpopular hot take: I am strongly of the view that it is NOT the responsibility of small-time writers and theatre makers to provide paid work for illustrators. I make maximum use of freely-available images, and I make no apology for doing so. If I’m going to make a lot of use of a free-to-use image or music, I will usually donate. The image I’m currently using for The Rats in the Walls was an AI-generated image provided by the venue – I make no apologies for that either. (I did check I wasn’t stealing anybody’s art though.) My budget is tiny, and I cannot be reasonably expected to pay someone a significant amount of money when I have an alternative that someone allowed to be used for free. You might have a better moral case against a bigger-budget production going cheap on artwork.
But here’s the thing – if I had the budget to hire an artist, I would. (You can hold me this this.) AI or no AI, there’s a lot you can do with free-to-use artwork if you’re tech-savvy, but bespoke artwork from paid designers is better. On the rare occasion I’ve been fortunate enough to be part of a bigger-budget enterprise, I’ve been amazed by the clever ideas that I never thought of. One thing’s for certain: a computer wouldn’t have come up with those ideas. Believe me, bespoke artwork is close to the top of my wish list, if any project of mine has income to sustain it. And as for bigger-budget productions, where it’s a case of can’t pay instead of won’t pay, you’d be a fool to go cheap on artists. That doesn’t mean they don’t – more on this in a moment.
As for whether AI will be a viable substitute for a creative writing any time soon – forget it. Now, I could write at length about what it can and can’t do, and I may well do so elsewhere, but the fact remains it can only use the source material used to train it. Ask ChatGPT to write a plot for Harry Potter fan-fiction, and it will probably do a reasonably decent job – that is, by definition, derivative of previous Harry Potter fiction. But I’m getting familiar with how language learning works; even if future AI tools could master disciplines such as scene structure, I just don’t see how it could come anywhere near the level of originality needed. At best it might be a useful tool for, say, writing mock Shakespeare or dialect. Might this threaten the income of people who can emulate classic authors or local language quirks? Maybe.
But I’m firmly of the view that technological advances have delivered far more benefits than problems to grass-roots artists. Digital cameras has made a craft that once needed a steep learning curve vastly more accessible. Professional photographers have not been driven out of business, because people who know what they’re doing take way better pictures than a random with a smartphone. Video editing used to be a vastly laborious task – now, it is in reach of anybody willing to learn. But whilst video editors in high-spec studios no longer have the monopoly on this trade, experience counts for everything. Even before the current AI craze, editing in Photoshop has been getting better and better and better, but the jobs went to the people who know what they’re doing, not just anyone with a Photoshop license. At worst, it threatens the careers of those who don’t move with the times. But how many creatives get by without moving with the times?
In summary, I’d like to conclude that creatives should have nothing to fear from AI technologies. AI technologies can help you in a craft, but AI will always do a substantially worse job than a human. But I can’t conclude this, because it falls down one basic assumption.
It assumes that AI doing substantially worse job is considered a problem.
Why AI might be a threat to artists after all:
The issue is that artistic merit is subjective. There’s nothing to stop you saying AI-created art is better than human-created art – as long as you’re prepared to move the goalposts far enough. And the arts are no stranger to the moving of goalposts.
There was a notorious case late last year when a Twitter user ran an AI improvement on this image: Edward Hopper’s 1942 painting Nighthawks:
“Sonch” listed all the shortcomings of this manually-created image. Terrible composition. The subjects are too far away from the setup. Nothing at all in left part of the image. You can barely see what’s going on. Let’s run one AI tool to get an image description and another one to generate a new image from the text. ta-dah! Now it’s a far sunnier inviting place. Look:
This got dunked on by everybody, but it turned out to be a hoax. However, as well as all the scorn, the “improved” image got over a thousand likes. It is not clear whether these were from people in on the joke, or from Hyperloop-loving techbros – I have a horrible feeling it was the latter. But what it did show was how easy it is to make subjective judgement in favour of whatever requires least effort.
Hoax or not, AI-generated Phillies remake will not be going to the Louvre any time soon; it was far too brazen. No amount of techbro-speak or pseudo-intellectual pontification is going to win the public away from a beloved piece of American art history. Nor is the art gallery-going public going to be swayed by the insight of an anonymous Twitter user with 5,000 followers. But this wouldn’t be the first time an artwork has been created with infinite acclaim for zero effort. How much praise has been heaped upon Damien Hirst for the fucking Spot Paintings? Would it still have been lauded as a work of genius without an army of sycophantic art critics saying so? I have a horrible feeling that had a well-known “contemporary” “artist” applied the same AI treatment to something not quite so well known, it could have got a lot more traction.
However, let’s be fair to contemporary art circles. This is hypothetical. For their faults, they have not yet gone down the route of lazy CGI. It’s also a niche interest with minimal implications for most artists. The real threat is in art for wider consumption. And there, the precedents are more concerning.
Take films. There’s been a lot of criticism over the use of AI as a substitute for actors. Crowd scenes have particularly come under fire, so badly edited that you can make out half a person in the final render. But why should they care? The big studios have long since cultivated an audience who cares for little but special effects. Formulaic plots are at worst a non-concern, at best a plus. Sequels, reboots and remakes are on the rise, original films are on the decline. Should we really be surprised that nobody minded AI-generated crowds in Michael Bay’s Big Explosions With Fragments Flying Left And Right VIII?
The problem didn’t start with AI either. Technology has historically done wonders for music, but it can also make it worse – if you can convince the mass consumption market that an keyboard demo with autotuned vocals is more fashionable than something that required effort and talent. Don’t get me wrong, there are some brilliant sequenced tracks out there. But one thing that has been sidelined from the charts is bands. (We still have singers with bands, but bands as the artists in their own right have virtually disappeared.) I believe pop music is a lot poorer for it. I also think it was convenient for marketing executives to convince the mass consumption market that bands weren’t fashionable any more. Musicians can become indispensable, musicians can talk back, musicians’ get artistic ideas that don’t align with making as much money as possible. Far better to streamline it all to identikit backing tracks, created by interchangeable – and replaceable – music techies.
The point is that AI-created art works best when it’s derivative work of formulaic tropes. And popular culture is already jam-packed full of this. The people responsible for marketing this have every interest in entrenching it further. Crappy pop culture is far from the only offender for redefining “great” art as whatever effort-free venture they’re already doing. At one end we have whatever guff the Tate Modern is fawning over this month (and woe betide anyone who calls it a Dulux Colour Chart); at the other end we have the amdram syndrome of redefining the greatest theatre as however they’ve always done it (and woe betide anyone who thinks there’s more to life than learning lines and always facing the audience).
What is most ironic it that, for writers in particular, the system has set us up for a fall. As I said, AI is good as derivative formulaic operations and bad and original creative operations, and the craft of writing is firmly in the latter. That doesn’t stop a lot of people in high places trying to make it the former. What breaks my heart about films and TV programmes that get panned is that the terrible writing is rarely the fault of the writers. They are almost always teams of plainly competent writers – who wouldn’t have got jobs without proving their worth – whose roles have been reduced to cogs in a machine. And if the people on high set terrible briefs, you are powerless to resist. Creativity and originality are driven out of you, derivate formulaic writing is now your fare instead. In other words, exactly the sort of job that can be rendered obsolete by machines.
The writing industry has been far too complicit in this. I know I keep ranting about this, but you know there’s something wrong when the most fantastic opportunity lauded upon aspiring wordsmiths is writing for Hollyoaks. Maybe it’s a steady job. Maybe it pays better. Maybe it’s even the best route into becoming a lead writer and have real career as a lead writer where you do realise your own vision. But when there’s so many talented writers trying to get their voices heard, heaping the most praise on the roles with the least creative freedom is such a let down. Now it’s asking for trouble too. If writing teams really do find their roles supplanted by AI Soap-bots, it will be the writing industry who made this possible.
AI is not a threat to artists’ livelihoods. It is greedy uncreative producers who are a threat to artists’ livelihoods. But AI is making it an awful lot easier to do their worst. AI may or may not develop to the point where it can take over formulaic derivative work – if it doesn’t there’s nothing we can do to stop it. But we can stop arts jobs being formulaic and uncreative in the first place. We should wise up before we fall into a trap of our own making.
And if we can’t see that. I give up. Let’s skip this stage and go straight on to Skynet.
Chris Neville-Smith's blog on theatre 