Can AI produce anything that isn't slop?

The internet is undoubtedly full of AI slop. From LinkedIn posts cheesier than a Stinking Bishop to Amazon books created from a single prompt, naff YouTube videos, and fact free Facebook images designed to enrage your racist grandparents.

The question of whether anything created with AI can avoid becoming slop or at least sloppier than it would be without it is more interesting. It relates to a lot of quite deep ideas about how expression, creation and writing overlap with what it means to be human and what we value in society.

Technical vs creative

A big divide for me in AI use comes between technical and creative work. For technical work, let’s take medical as an example, the most important thing is output. How is it helping someone or more often people en masse cheaper or more effectively? It’s hard to argue that lots of technical things shouldn’t maximise efficiency (though as a separate point I do worry about mass burnout as we complete tasks and switch tasks so much faster without actually reducing workload.) So when it comes to digesting or writing technical work with a view to say making better prosthetic limbs or widening access to cancer treatment, I don’t think anyone would argue we shouldn’t optimise research and dissemination of that research with AI.

At the other end of the spectrum, for creative work, output isn’t really what matters. The speed at which a book or a song is written doesn’t matter to the person reading or listening. A novel is more than the sum of its parts. It is not just the moral of the story, the summary of the plot, or the eventual character arc. It’s the journey of the thing, the rhythms and cadences, the slow accumulation of character choices, the negative spaces (what’s not shown or the perspectives not looked from) as well as the positive ones. I have a lot of extra thoughts on why AI shouldn’t be used to write novels, but that’s a question for another blog.

The capacity of artists to create in their lifetime isn’t really an issue either. Whereas for technical work, you want your genius medical researcher solving as many problems as they can, that doesn’t apply to great artists, who can already produce enough in a natural lifetime. In lots of ways reaching the end of someone’s catalogue is a good thing because it makes you go and explore other artists. If someone uses AI to immediately create 1000 novels in the style of Robert Harris and 1000 paintings in the style of Andy Warhol and that was all they chose to read or put on their walls for the rest of their lives they’d be missing out. Art is meant to enable us to watch and read and hear different things rather than create a homogenised experience. In fact it may already be making users lonelier.

And, finally on output, accessibility is a false flag. Our society has already created ways of widening access - libraries, radio, free exhibits etc.

The value of originality.

The other side of the output coin with creative endeavours is that even if we leave aside the value of it, AI can’t actually guarantee it. What humans value as art is so idiosynchratic that AI can’t just automatically produce a bestseller or a symphony that connects with people. Humans have always valued new things. Shakespeare coined new words and George Lucas coined new worlds. Jazz took hold because the unusual beat caught people’s imaginations. Some maverick once decided to deep fry a Mars bar.

LLMs, by design and definition, default to the safe and conventional middle. All its word steaks get done medium-well. Its musical curries are mild. Everyone starts to sound the same, and that sameness compounds over time as content gets fed back into LLMs. To use a less explicitly creative example, I did battle with Google’s AI over the weekend while choosing a paint colour. Useful and clever in lots of ways (it encouraged me to think about orientation and light and furniture colours) but it kept defaulting to the safest and most popular colours. Scale that up societally and we cook up or paint our walls with a giant vat of beige. More people eat the same, dress the same, read the same books, have the same thoughts about the same things. Boring.

An aesthetic sense of time

I would also argue that humans tend to have quite an intrinsic sense of the value of time and an ability to identify what took time. Beyond nostalgia, one of the reasons the LOTR films have endured so well whereas The Rings of Power doesn’t really connect is the level of thought and detail that went into the production. Miniature models, handmade armour. We can still tell the difference between those and CGI, between that and AI films, and that difference computes into an opinion. Rather than meaning less time for other elements, taking the time to get those things right tends to correlate with spending more time on other things from screenplays to scores.

Even if it could

If new models got closer to genuine original thought or could recreate more perfectly the sense of painstaking creation, would it matter? If I get more joy in my hard earned leisure time by playing a ‘better’ video game made 100% by AI or if looking at an AI painting on my wall offers me more stimulation or serenity than a human painted one, should that matter? I suspect not to a lot of people, but I would argue it should.

To paraphrase Alan Bennett, one of the great moments in reading is when it feels like a hand has come out and taken yours. That’s true I suspect with other forms of art too - we value that another human has made us feel something. That connection becomes cold and metallic when it’s with a machine. And what’s more, because we can only consume a finite amount of art and culture, any connection we get with a machine is by nature replacing another connection we could have had, so divorcing humans more from the world and people around us. If we can all create our own perfect things, linger in our own microcultures, we don’t have to share things or recommend things or gather together for cultural reasons.

There’s also a question of how the AI models would have got to that creative inflection point, which touches on topics of ownership and theft. In say academic research, the goal would often be to produce work that gets read and built on by others. The researchers have often (though not always - a separate conversation) been compensated for that work up front rather than relying on reads / listens / watches to earn royalties. AI companies scanning books so that other people can reuse their phrases, characters and plots with less effort is very different. The person creating reams of Warhols or Harris' through AI is not paying the artist or their estate a penny, a dollar or even a vietnamese dong for that work.

The middle of the spectrum

Most things of course sit in between technical and creative and that’s where the lines get blurry. Should a brilliant restaurant get penalised for writing website copy or producing a poster with an LLM? If a recruiter fills their LinkedIn page with AI slop but happens to find great prospects and secure those prospects great salaries, should we think less of them?

While I would argue it matters less than in an openly creative pursuit, personally it does negatively affect my view of a bar or a hairdresser or whatever else that uses AI images for their advertising. Partly that’s aesthetic (I just think a lot of AI has a very colour saturated look that’s naff), but it’s more that I see it as a sign of laziness. I’d prefer to support a business that’s put more work in, that’s thought harder about what makes it unique and has taken the time to communicate that to me. Is the recruiter relying on AI slop getting lazy in other parts of their business, cutting corners in search of a quick deal?

Process slop vs Output slop - the cognitive impact

Perhaps most importantly, there’s a question of slop of process vs slop of output too. I used Canva AI to create a LinkedIn post for a company recently. Design wise it’s cleaner than I could produce solo and it took less time. Output for the client is better. But the process for me was boring and sloppy. It required much less creativity, much less brainpower, much less learning than if I’d had to do something lower tech. I don’t feel proud of what I’ve created in a way I might have done had I spent longer on something lower tech in previous years. I don’t want to go home and talk about it with my wife. Any skill I’ve learned in promoting isn’t interesting enough to offset the loss of skill and loss of diversity of my other skills over time. And be assured that it is costing us skills. There is early evidence (albeit sometimes with small samples) in writing and other fields that AI usage not only produces less brainpower in the moment of its creation but has long-lasting impacts on memory and engagement, creating cognitive debt. It is also reshaping how we think, reason, interpret culture, and communicate with each other. It might even be neutralising our response to questions and making us more middle of the road emotionally.

To go back to the Canva example, I also might have had to pay a designer to do it five years ago or a company might have had one in house, so someone else has lost work to the tech.

Clearly there are situations where the equation is slightly different, where a blog or image involves collaboration between multiple experts, where AI is used in a much more engaged and engaging way. There are situations where being able to use AI might mean a business has fewer hurdles before to launch and can go on to do great things down the line. The sloppiness of those is mitigated to a much greater degree than a single prompted blog or poster or song.

Obviously AI isn’t the only thing that’s ever shifted the way we approach creative or semi-creative tasks. The CGI vs models and handmade film props is a good example. And people have lost jobs to tech through history. But I would argue AI is both more fundamental and widespread a change than ever, and one that is atrophying at a stroke more of our creative muscles.