My non-fiction isn’t “content” either.
I agree that there’s a lot of terrible business communication out there. It will not be improved by handing it off to the slop generator.
(Our support queue is starting to get LLM-written queries. Probably just sending us the prompt the person started with would be more useful.)
One of the thing LLMs are actually sort of useful for is sombing through the tombs of untold numbers of forgotten trees in search of the answers to arcane and obscure lore…. And occasionally return a correct answer.
In all seriousness, it plays right into the current societal fade of wanting everything now….
As a lark, why don’t you take one of the more common support requests you get and feed it to google as a natural language request for help to see what it spits back at you. The results might be enlightening.
Except it doesn’t! Just send us an email! I don’t understand why dictating a prompt is supposed to be faster!
Oh when a user starts with a prompt, and replies to it, the agent will at some point offer to help the user contact support with a summary of the agent’s (often flawed) understanding of the problem. Thus relieving the schmuck from having to think about it further until the other-other brain fixes their problems for them.
I used to work IT in the 90’s the ‘unskilled operators’ back then were just as bad.
Fine.
Then put co-authored by ChatGPT on the cover. If it’s that great and there’s nothing to hide, everyone should just get over it. I mean, if you didn’t actually write all of it, or edit it…
(with tongue in cheek!)
Well that’s certainly a sad start to anyone’s morning: writers submitting AI-generated support requests for a writing app. 
Printing didn’t end oral traditions (campfire talks, barroom stories, or library story hours). Photography and cinema did not end handcrafted arts. AI will not replace writing by people; it might replace or displace some reades though.
Well, I can’t say I remember the last time I did any of those things, but I can remember the last time I saw words in print.
I also can’t remember the last time I partook of the handcrafted arts (I own no oil paintings, and it’s been a good twenty years since my last actual real painting)… but I can remember the last time I went to the cinema (it was yesterday, to watch the excellent Coyote vs Acme for the second time) and the last time I took a photo (it was about 3 hours ago).
So, these things might not have destroyed their predecessor arts, but they certainly have had an impact.
Something about 90% perspiration, or something like that, comes to mind (or something like that).
1% inspiration, 99% persperation
Many are making comparisons with inventions that we once feared would threaten or even end established art forms. For example, photography would replace painting, recorded music would replace live performance, cinema replace theater, television replace radio, and digital tools replace traditional forms of illustration and design. The argument is that these technologies generally transformed existing art forms rather than replaced them, and they often ended up coexisting.
They’re forgetting that AI represents an unprecedented technological evolution. Nothing in history even comes close. AI is not a single or a few inventions challenging one particular art form. It’s an entire ecosystem of machine-made art that spans multiple media and encompasses myriads of traditions, professions, and activities.
The combination of its scale, speed, versatility, and ability to generate outputs across numerous art forms, is unprecedented.
Previous inventions typically changed how people created, distributed, or experienced art within particular fields; AI has the potential to perform, imitate, and mass-produce creative work across many fields at once. Although historical comparisons can offer limited context, no earlier technology matches the breadth, speed, and direct threat to human creative labor posed by generative AI. Its impact must be understood on its own terms, rather than treated as just another episode in the history of artistic innovation.
I’m a musician, composer, music producer, illustrator, painter, designer, art director, and writer. I’ve worked in all of these fields in one capacity or another, and nowadays I do so at a low level on a hobby basis, from the unlikely vantage point of a disability pension.
I do not like AI in any creative capacity whatsoever. As a tool to aid creation of art without ever contributing with creativity itself, sure. But it should end there, if I had a say.
So, here’s what I’m suspecting might happen, sprinkled here and there with wishful thinking:
Everything related to art, with the exception of live performance, will eventually be more or less replaced by AI. Because, compared with human labor, it costs next to nothing to produce. And money has always been the final arbiter of what survives.
Eventually, live performance will be more or less replaced by androids. This will be a fad. People will marvel at it for a while, and then, as tends to happen with technological gimmicks, eventually grow bored. At some point, they’ll realize that what they miss is the human connection.
Human-made art will make a comeback. Not because it’s necessarily better, but because people will once again discover that they miss the presence of another human being behind the work.
Eventually, humans will become symbiotic with AI and robotics, and the distinction between human-made and machine-made art will begin to dissolve. And greatly confuse.
People will keep creating, just as much as they do today, because for the most part, people create for reasons that have very little to do with whether anyone is willing to pay them for it. The difference will be that most of them won’t make any money from it. Those who can perform their art live may be the fortunate exceptions.
Eventually, all of this will become completely irrelevant when the planet no longer has the capacity to sustain human life. Art will die along with us. At that point, dinosaurs will make a brief comeback, only to be smacked out of existence by yet another really annoying asteroid.
But until then, human-made art will survive.
AI will eventually be able to perfectly imitate the work. Androids will eventually be able to perfectly imitate the performance. But the reasons we make art – and the reasons we seek it out – will remain stubbornly human.
The reason art forms have survived previous ‘revolutions’ (and there’s an argument to be made that most have actually only survived in a decimated form) is because historically people made art because they wanted to make art. It was the creation that they loved.
If your joy comes from the act of creation, you are unlikely to be interested in automating it.
If, however, you are interested in creation merely as a means to get the output (which is almost all the people in the audience for created works) then automated processes that cut out the pesky passionate and emotional creators is not only tolerable – it’s a bonus.
So, you’ll always be able to play guitar in your bedroom. You’ll always be able to take photos for your personal enjoyment. You’ll always be able to scribble a poem in biro on the back of your maths homework. No one can ever take those things from you… but the market to do so professionally? Welcome to the shitty side of an Industrial Revolution.
Have you ever toyed with any of the AI frameworks out there? No need to pay ChatGPA or Google’s Gemmi…. You can download the open source Ollama app, and from there download one of the many models of LLM to your local computer to explore what it can do. You might find it illustrative to learn how much effort it takes to get it to output what you are looking for.
I have. I have very little artistic ability when it comes to drawing/painting, and have reasonably useful talents at logic and description. I’m having to ‘re-learn’ how to talk in order to prompt correctly. Asking the wrong question, or looking at the returns in the wrong ways will destroy the value of the results.
And that is before you start seeing the way the models are ‘trained’ to reply to conversational prompts with politeness that feels a fake as the rest of it. Unearned superlatives, highest praise, sycophantic agreement. Can you think of anyone that might resonate with? Maybe the people that wield the money like their own personal army? Maybe the people that have egos that enjoy being stroked… Either way, it’s an interesting study managing expectations gone wrong.
I’m serious, find one and play with it, and I do mean ‘play’. The more you play with it, the more it will cease being a boogeyman of replacing the artist (who have always been starving) and more of yet-another-tool-to-create-yet-another-form-of-art.
I’ve used ChatGPT extensively for about a year on projects unrelated to art, and getting it to do exactly what I want isn’t easy. I’d say it’s rather painful at times.
However, getting it to create reasonably good art for me is a piece of cake. But I think what’s relevant in the debate isn’t how difficult it is – or isn’t – to make AI create certain kinds of art for you today. It’s how easy it will be tomorrow.
The struggling artist will always be struggling, for better or worse. For better, I think, in many cases.
I’m mainly talking about successful artists who are receiving fewer commissions as AI-generated work becomes a cheaper substitute in their markets. Their styles are also copied by AI. You can tell AI to create a Wallace and Gromit-style illustration, and that’s what you’re gonna get. In minutes. Copyright laws have so far not been able to keep up with the speed of these developments.
I think AI will, for the most part, never be a tool to create another form of art, but rather a tool to steal other artists’ art. Because that’s what’s at the very core of AI-generated art: infinitely imaginable ways of reusing what artists have created. And as soon as AI creates something for you, you’re not creating. But that’s just my opinion, not a fact; far from it. But alas, that one is an endless topic to debate.
The assumption that “tomorrow’s” genAI will produce better results is unlikely to hold. We’re already well past the point of diminishing returns, and there’s some evidence that genAI output is actually getting worse.
But isn’t it premature to conclude that meaningful improvement has ended? Diminishing returns aren’t the same as an approaching endpoint. Perhaps future improvements will come not just from making models bigger, but from making them better at reasoning and checking their own work, i.e. an AI-generated painting of a guitar player will no longer have seven fingers and guitar strings running through their hands, like today.
Just playing the devil’s advocate – I’d very much like your statement to come to be.
It sort of is, given that money, electricity, and water are not infinite resources.
But what does “check their own work” even mean? They don’t have anything to check it against other than their training data.
Indeed, finite resources may eventually impose a practical limit. But diminishing returns still don’t mean that improvement has ended; they mean that each additional improvement becomes more expensive. Progress may come from better or more efficient methods, not simply from using more money, electricity, and water.
Fair point. By “checking their own work”, I don’t mean that a model can just look at its own answer and somehow know whether it’s right. I mean giving it ways to actually test what it produces, whether that’s against specific rules, by checking an image, using external tools, or having something else evaluate it. In the guitar player example, it could check the image for obvious anatomical problems and fix them before showing it to the user. That doesn’t seem to happen reliably at the moment. They mostly seem to be generating things based on patterns and trying to interpret what the prompt is asking for.
To add an example, I’m a trumpet player, and these AI-generated images are mechanically and anatomically wrong on so many levels that it would be hilarious if I didn’t find AI as utterly creepy as I do. But I suspect methods will be developed to fact-check them against reality.
If there were a way to do that algorithmically, it wouldn’t be hallucinating in the first place.
Human artists “test” their images against their understanding of physical reality. (Or against the image in their head, for non-representational art.) GenAI has no “understanding” of anything.
Perhaps I didn’t express myself very well, but I wasn’t suggesting that current AI has human-like understanding or can reliably check its own work. I agree that its lack of those abilities is a major reason it produces these hallucinations and physically weird images.
My points have been about what might be possible in the future. What AI can do today was unimaginable not long ago, so who knows what it may be capable of tomorrow, with the help of future inventions. We often don’t know how new technologies will evolve, and they often develop in ways that are difficult to predict.
My views are only guesses – frankly, they’re also shaped by my fears and dislike of AI.
This has gotten more lengthy than I intended; I wrote my original post hoping to perhaps contribute something, and leave it at that. English is my second language, so writing each reply takes me much longer than it would in my native language, simply because understanding someone’s reasoning in English and expressing my own in English is much more challenging. But thanks for the discussion, cheers
