Why is AI Taking Over Creative Industries?
We’re still pressing buttons in an office all day while software engineers work to automate Netflix scripts
I didn’t know anything about AI until earlier this year.
That’s not true. Of course I’ve always known what AI is, I grew up on science fiction. What I mean is I didn’t know anything about what we’re currently referring to as AI, which isn’t AI at all. It’s not actually intelligent. It just recognizes and replicates patterns. It’s basically a more robust version of the autocorrect feature on your phone. It’s Clippy on steroids.
This stuff first came to my attention when everyone on Twitter started posting images they made with Dall-e, an AI image generator that can purportedly draw anything that you describe textually. I’ll admit to having a lot of fun with it for about an hour or so before it occurred to me that I already have the power to perceive images of descriptions I come up with and it’s called using my imagination. And my imagination is usually a lot more accurate and faster.
The truth, as it later became widely known, was that the software wasn’t drawing at all. It was taking images from the internet that were already created by human beings and mashing them together to half-assedly match the descriptions being fed to it.
Very soon the hype train moved on to ChatGPT, a language model that will answer any question you ask it, which is the same thing Google does already, except ChatGPT answers it in full sentences like a person would. Almost immediately right wingers started having panic aneurisms over the fact that they couldn’t make it say the N-word, so it was obviously a communist.
It is legitimately an amazing technology, don’t get me wrong. While it might not understand things per se, it can nevertheless interpret prompts and environmental stimulus in ways that are far and away better than older technologies like search engines. It’s much faster at pattern recognition. It’s not hard to come up with thousands of potential applications. Even faster if you ask ChatGPT to come up with thousands of potential applications.
So why are we using it to replace artists? You know, to do the one thing it’s worst at?
When the Writers Guild of America went on strike last month, one of the biggest grievances besides, you know, wanting to get paid for the work they do, was that they wanted assurances that their profession won’t be replaced by AI. It’s a reasonable fear. Such things are already happening. Buzzfeed is written by ChatGPT now. First they came for your listicles and we said nothing.
The anxiety about automation has been ingrained in us ever since the Luddites smashed up the textile mills, but there was always an underlying assumption that the creative industries were going to remain the domain of human beings. That, in fact, the automation of most gruelling manual work would free us up to invest time in the artistic pursuits that make us uniquely human. So why is Amazon’s Kindle store being flooded with books written by an algorithm? Why are we trying to get AI to write TV episodes before we have a driverless car that knows not to run down children at a crosswalk?
Like most problems, by which I mean early all problems, the answer has almost everything to do with money. ChatGPT is free, human writers are not. And when you combine bottom-line profiteering with a soulless and almost total inability to comprehend human joy, you can kind of see where this is headed.
Along the same track as what I wrote about last week, a lot of people are just unable to parse the idea that people should be allowed to, or even have the ability to, enjoy the work that they do. So you wind up getting frankly bizarre takes, like how we could make a lot of money out of the fiction industry if we simply took the burden of actually writing stuff and reading stuff out of human hands and give it to a computer and then deliver to consumers what we suspect they really want, which is having realistic personal/sexual conversations with Harry Potter.
But as Twitter pointed out, the appeal of reading fiction isn’t that it’s some clunky and antiquated intermediary step between paying money and getting endorphins. The creation is the magical part.
AI, or what we’re calling AI, is technically able to write a story. But only in the sense that McDonald’s is technically food. The people in charge of McDonald’s aren’t chefs, they are capitalists. Similarly, the problem here is that the people in charge of Netflix aren’t filmmakers, they are capitalists. The people in charge of Simon and Schuster aren’t novelists, they are capitalists.
Charlie Brooker once tried to generate an episode of Black Mirror with ChatGPT. And props to him because that’s actually a really amazing idea that would be completely in line with the theme of the show if he’d gone through with it. But he scrapped the idea because it was just a shit script—all it could do in response to his request was take a bunch of other Black Mirror episodes and just mash them together. Because, again, that’s all it does.
ChatGPT has no creativity of its own. Human beings are the engine that makes it run, because it feeds on the creative output of real artists for the raw materials to assemble its own “creations.” It just becomes incestuous and derivative sludge when its artistic gene pool is too small, like Brooker’s sad Frankenstein of an AI generated Black Mirror. But the other issue is that scraping the hard work of real artists off the internet, disassembling it and reassembling it into something you’re disingenuously calling an original work is something that a growing number of increasingly angry artists understandably believe to be a form of theft.
Recently a bunch of start-up techbros got grilled to a golden brown over the announcement of a new fiction-generating AI product they’re calling Sudowrite. The outrage comes from two angles—one, that they’re basically seeking to turn the writing industry into one giant beefed-up game of madlibs, and two, the raw data that they scraped for their project came undisclosed from an earlier project in which they asked writers to send them their unpublished manuscripts, for which their AI would generate loglines and synopses for them while deviously devouring their creative energy for regurgitation into one-click robo-books.
That is the key difference, though, between trying to automate creative industries as opposed to automating physical toil. Artistic works need a human origin to function in the first place. You can automate a road vehicle completely (though they’re having a lot of trouble working out the kinks) because the concept of moving an object from one place to another needs no human insight. The machine is moving the human but the movement begins with the machine.
When you’re talking about art, whether it’s writing or filmmaking or visual art, the movement always, always begins with a human. It’s literally impossible to do otherwise. You can’t just feed the exact same data back into a machine over and over again and generate novelty, and it is novelty that sits at the heart of the meaning and purpose of art, and the joy it gives us.
People accuse Marvel movies of all being the same film re-skinned over and over, but brother, you haven’t seen what it would be like if they actually were the same movie the way they would be if the machines took over.
Unfortunately those impatient for the quick buck are happy to risk killing creative industries in an effort to make it work. For example, the science fiction magazine Clarkesworld is one of the highest profile publications that’s had to halt manuscript submissions while they try to figure out how to move forward in this new world. Why? Because the volume of manuscripts being lobbed at them that were generated entirely with ChatGPT has become impossible to handle.
The problem is not that they, or readers, cannot tell the difference between bad algorithmically generated fiction shat out by a glorified pocket calculator. The problem is that they don’t have the manpower necessary to drudge through it all in search of the good stuff. It used to be that the slush pile in publishing houses, while burdensome, was manageable due to the fact that it still took a lot of time and effort to churn out each piece, bad or good. If an author can pump one hundred words into a ChatGPT prompt and walk away and make a coffee while it puts together fifteen terrible short stories, that’s really bad for quality fiction magazines, and really bad for real human writers. People like me.
A world in which real art is replaced by sterile computer-generated simulacra to keep us placid and entertained while we drudge at our manual labour ironically resembles the exact kind of imaginative fiction prompt we stand to lose. And you can bet that whatever kind of stories the machines generate, they won’t include AI as the bad guys. The one benefit is that this is the only way we’re ever going to stop the Terminator sequels.
Bravo S Peter Davis! We need more creatives writing about the gen AI impact on our work. And getting those articles out into the world. We need to keep talking about it, and insist on retaining our human right to create, because it is a HUMAN RIGHT. It's a human right that's being taken away not by AI, but by other humans.
Put simply, generative AI taking over the creative fields is really about one human telling another human they are no longer valued—because they weren’t truly valued to begin with. That's what it boils down to. We are the auto-immune disease.
https://themuse.substack.com/p/we-are-an-auto-immune-disease
p.s. A little more context about the Jesus Christ painting at the top of this post... on the left is a 1930 fresco of Christ by Elías García Martínez, titled Ecce Homo and on the right is the result of an 81-year-old woman who had been permitted to try and restore the painting, and of course failed miserably. Kinda like what AI is doing to our human art works.
It's also funny in a dark way that these LLMs and dall-es are being used to gobble up the jobs with the least amount of personal profit attached to them, in the present day "entertainment industry". Nobody's further from the next dollar than the writer or musician or artist, after all. In any case, the flood of rock-dumb profiteers trying to use this stuff to crank out low-effort "content" are going to ruin it soon, as their models begin to poison each other's datasets, either accidentally or on purpose.