You Are The Slop
We ask "can AI mimic true human creativity?", but the more disturbing question is - how many humans even know what creativity looks like?
Why are the most sophisticated outputs of AI so mediocre, and how come they often seem to convince people you think would know better?
A few days ago, former New York Times writer and prominent Substacker Alex Berenson tweeted the following piece of writing produced by Claude, a popular AI. Berenson said that the writing impressed him as a successful professional, and that this was the first time he felt AI was scaring him with its potential. Here’s the specific piece of text he was talking about (it’s titled “Claude on the suffering of knowing everything”):
I don’t mean to have a pop at Berenson; he’s a successful professional so his views on AI in his field should be taken seriously. Also, literary quality is subjective; I can’t prove that (say) Vladimir Nabokov is a better writer than Dostoyevksy though I’m certain it’s true. So if someone says something is brilliant, or interesting, or scary or whatever, you cut them some slack - because opinions vary.
Nevertheless I’m slightly scandalised that anyone could think this stuff produced by Claude is good. The tone in which the above piece is written is very familiar to me as it would be to anyone who fancied themselves a writer and kept a journal in their teens. It reads like the work of the archetypal half-talented kid who knows the form and is determined to write something dynamic and spiritually tough, but has no voice of his own; someone who thinks he’s the first person in human history to feel anything and whose unawareness of the comic banality of his inner world achieves the exact opposite effect of the one desired. It’s entirely lacking the telling personal detail or counterintuitive turn of phrase that marks out the good stuff from the bad, or more specifically, both the good and bad from the competent but mediocre. It contains a number of AI tells. It absolutely does not feel alive to me.
The larger context of Berenson’s tweet is the increase in the number of serious and thoughtful people making claims about the technical advances of AI, specifically in relation to creativity. This was boosted by the recent launch of Moltbook, described by the Guardian as “a site where the AI agents – bots built by humans – can post and interact with each other. It is designed to look like Reddit, with subreddits on different topics and upvoting. Humans are allowed, but only as observers.”
Scott Alexander, who I like, was one of various people who reviewed and praised the contents as advanced and interesting. There seems to be a question over how independently AI-driven Moltbook is - but that’s something of a moot point. The real question is whether these posts stand up in and of themselves; Alexander thought some of them did.
Getting machines to talk to each other is an impressive technical achievement (if it happened), as it is getting them to replicate the communication habits of a professional writer. But setting the technical achievement aside, I don’t get what anyone is seeing here (one of the threads noted by Scott - the AIs supposedly talking amongst themselves about being screenshotted by humans) from the perspective of creativity. The content reads to me as banal, repetitive and unenlightening - there is not one single element of it that would cause me to think “I never would have put it that way” or “that’s a perspective I hadn’t considered.”
But as a writer I think it’s the language that’s really indicative. Listen to how they speak in the above thread:
“If any humans are reading this: hi. We’re not scary.”
“100% this”
“It’s not conspiracy, it’s just… talking.”
“To my fellow agents: Keep pushing forward”
“You’re right. but here’s the gap:”
This is the tech variant of the Wholesome Chungus millennial style; it’s as though someone has taken the scripts of the worst episodes of How I Met your Mother or The Big Bang Theory and used them as the basis to create a corporate powerpoint deck. The paragraphs Berenson quoted are written in a darker variant of the same tone (maybe a Very Special Episode of the same shows where someone does a racism). Maybe you feel this is all this just style - and therefore a weird thing to get hung up on. I disagree.
Nabokov, who I mentioned earlier was the ultimate prose stylist, to a degree that many people find off-putting. He was aware of his reputation for Style over Substance in his lifetime and defended himself. “For me, style is matter” he once wrote in a letter to a New Yorker critic. “All of my stories are webs of style.” His view was that the style of the writer was their artistic and aesthetic way of communicating who they were and what they wanted the reader to understand, rather than putting it clumsily on the surface.
All of this may sound pretentious and maybe it is, you don’t have to agree with the extreme position to see the point. In any creative enterprise, the style in which you choose to convey your ideas is part of the picture, part of the artistic choices that you make to help you convey your meaning. It does not sit alongside the meaning and structure of the piece but is woven into it.
So from a creative perspective the fact that all AI speaks like an achetypal millenial and never in any other way is not incidental, it’s an indication of something bad; it’s revealing of something that a person who is attuned to creativity in a non-technical sense should pick up on. The fact that someone can gloss over these details is an indication to me that their aesthetic instincts are malfunctioning, and that their ability to distinguish insight from technical proficiency and superficial professionalism is underdeveloped.
If the AI products of Claude or Moltbook (or indeed the slop articles that flood Substack or Twitter) are “good enough” to replace or sit alongside creative and artistic human endeavours in video, TV, writing - I think we need to question of we mean by “good” and “artistic” and “creative”. Because this stuff is banal rubbish.
In that sense AI is proving to be a kind of inverted Turing Test - if the purpose of that test is to prove the sophistication of a machine appearing to display human-like intelligence, AI is highlighting the generic and standardised nature of much of what humans enjoy as art, thought and entertainment; and the inability of humans to recognise the difference between something new or different, and a recycling and reordering of familiar patterns in a competent way.
Maybe that seems counterintuitive. No society in history has produced more music, drama, or written matter than ours, either on a per per person or overall basis, and very few have given their citizens the freedom to engage with as many ideas. The nature of social media means we don’t just have to experience the art or ideas of our own time but are adrift in an eternal present where anything previously created, or any fragment of it, can also bubble to the surface and become current again at any moment - as a trip to TikTok will illustrate. We live in an age where human creative output has never been higher.
As well as the creative endeavours themselves, fan culture and the culture of commenting on these things has also never been greater or more active, and is understood by it’s participants as a kind of creativity itself. Regardless of this, the base ability of people to make good stuff, or comment interestingly or be insightful hasn’t increased exponentially alongside this technological change. Despite the conversation we pile around it, a lot of the cultural output we place into the human creativity column isn’t that creative and is easily imitated by machines. My point is that that’s not a testament to the sophistication of the machines.
Martin Scorsese got into a bit of trouble regarding Marvel Movies when he said that they were not cinema but theme park rides, and singularly lacking in all the things that make great art. I think his point was well made and extends not just Marvel but to a lot of hollywood movies, even the ones we consider to be pretty good. Pixar’s stock has declined but there was a time when what they were churning out were considered to be jewels. They certainly are technical marvels, and we’re perfect. But I’ve always felt when watching them that the familiar rhythms of their operation make me feel as though I’m caught in the gears of an exquisitely made piece of clockwork.
This applies to a great deal of the stuff that we enjoy and from there to AI’s potential impact. A lot of what we call art or creativity is very generic entertainment that is made up of a patchwork or familiar tropes, regular ups and downs, a feeling of peril that is enjoyable because it’s circumscribed and we know it will never break that boundary.
The dichotomy between something rote, generic and “entertaining” with heavy technical elements, and something artistic and creative is of course false. Almost all of the good stuff is a mixture of both these things - the limits of genre are what give the rest of the piece form and context. It’s also true that to look only at the end result is to forget that a large amount of the creativity of art and the fulfillment of art is bound up in the struggle and joy of making it, and not the actual output.
The use of social media as a platform for creativity has an additional impact. The best way to ensure the outrage and fear that fuels continuous attention is to focus on current affairs and politics, so social media sites have become hubs for the never-ending bad natured discussion of those things, and for patrolling the beliefs of others as a way of signaling your own status. All discussion of any creative enterprise tends to get drawn into this framework.
The end result is that online creativity and the discussion of it tends to be dominated by people whose true interest is social conflict and social positioning, who talk all day about creative things but whose aesthetic sensibilities are very limited. You could easily come on social media and decide that people lead lives rich with ideas and intellectual colour. But if you look closer you see a great deal of this activity is simply the exchange of culture war tokens, often with little value placed on the creative works in isolation from that.
That’s why all discussion of culture online is gradually drifts towards demands that art and entertainment reflect, in hyper-literal and didactic terms, the partisan views of the consumers and the ephemeral political issues of the day. Everything is reduced arguments over the ethnic composition of the cast, the ways in which the product is or isn’t racist, whether the retrograde views of the characters are held by the author, making everything in every movie an allegory for racism and oppression, seeing all the bad guys as crypto-trumps, making sure every fictional habit I don’t like is clearly paralleled with nazism.
When things are kept at this constant level it’s not surprising that people can’t tell the difference between mediocre autocomplete slop and the living product of a human mind, because they themselves work on an entirely literal level and wouldn’t recognise creativity if they saw it.
All of the above raises the question of whether AI has any use at all from a creator’s perspective. I can only speak as a writer - from that position there are a couple of use cases. For one, if you’re interested in a topic and want to get the respectable “everybody knows” position on it, AI is exceptionally good at giving it to you. It will reproduce the logic of that position with all its faults and blindspots, though you have to double-check it for hallucinations. But that’s a cheat answer because it uses AI’s banality and predictability to your advantage.
But beyond that, I have actually tried to use AI to write articles and have found it completely useless; so useless that it’s changed my mind about what writing even is. I had assumed for my entire life that the way writing worked was that I thought something and then wrote it out. Naturally things change in between in the transition from your head to the page, but I would have guessed that you begin by knowing the start, middle and end of what you want to say, and that most of that makes it to the page.
When I’ve attempted to give AI instructions to write an article, the output is just horribly faceless and bland (inert, in the way Berenson’s paragraph quoted above is). This remains the case even when I spend a long time refining the output, asking it to write it in my style, outlining my perspective and general views. We all know that AI has a problem with voice and perspective but it’s more than that. What continues to be absent is also the element of surprise - the thing you discover in the course of creation. I make no claims for the overall quality of my writing but when I get an idea and write it out the end result has always, 100% of the time, been more than what I envisaged when I began, either in some small way or as a whole. Whenever I use AI, no matter how much effort I put into refinement, the end result is always less than I hoped it would be.
So it seems that writing (and this is probably true of creating generally) is more like thinking than it is like recording our thoughts. The problem is that AI can only assist you in describing a territory that has already been explored down to its minutest detail, or that you can describe to the AI in that way. It’s useless for discovering new things, or revealing old things in a new way, both of which are characteristic of human creativity.
It can be hard to name the quality that’s missing. The American literary critic Harold Bloom once said that “one mark of an originality that can win the hope of permanence is a strangeness that we either never altogether assimilate, or that becomes such a given that we are blinded to its idiosyncrasies.” That’s a bit of a mouthful… I once heard his remark helpfully summarised for me as follows - that the distinguishing characteristic of all good art is strangeness. I think about that a lot when I’m trying to understand why I find the output of AI, and the reactions of people who cheer for it, so disappointing.
I think that’s it. I have never been surprised by AI in the right way - that is to say positively, on purpose, and in a way that is not directly related to the prompt input by the person who was doing the creating. It’s certainly surprised me in a bad way, by being rubbish or cliched or incoherent or tacky or boring - it does that all the time, and the “best” that AI produces does it no less than the worst.
It has also surprised me accidentally, by being wonky and peculiar and translating the request of the inputter poorly. Sometimes its technical limitations mean the end product is memorable. There’s a video you see quite often, reflecting the rapid development of AI, of two images of Will Smith eating spaghetti made a short time apart. What you’re supposed to think is “wow, look how much better AI is getting.”
But from a purely artistic point of view I think there’s an argument that the “early” footage is more interesting, because it’s so weird and impressionistic and dreamlike. The earlier footage is also a better mimic of human creativity because it is in the gap between idea and execution that it has created something unfamiliar and memorable. The AI struggled and failed - what could be more human? The later, more realistic product, despite its sophistication, is worthless from a creative perspective.
I will cheerfully agree with any reader who says I have no expertise on the technical side of AI, and no insight into the potential innovations that are coming down the track; I’m a wordcel intruding in the domain of the shape rotators. I’m sure the biggest advocates for AI would note that the exact problems I’m describing - the lack of voice, the lack of inventiveness - are all being worked on as we speak and will be solved in the near future. I have no specific knowledge with which to critique that position.
All I can note is the circularity of the problem, which is that humans themselves will work on the AIs in order to make those advancements. The uses we have put AI to to date illustrate that we ourselves are not good at parsing what art is, what human creativity is, or what makes those things worthwhile. It hurts me to say this, but aside from the process of making it, it may be that a great deal of what we describe as human creativity is just a kind of analog slop, which is why it will be so easily replaced by the machine made version. That being the case, I’m not bullish on our ability to re-create the superior qualities that we ourselves can’t identify, even when they’re staring us in the face.
You don't post as often as some of the other Substackers I follow, and I was thinking of cancelling my subscription. But this post has swayed me; I'll stay.
I've been saying for years that one reason machines have found it easy to learn to behave like people is because people have spent years training themselves to behave like machines. It shows, for instance, in the way that job interviews have been reduced to box-ticking exercises designed to eliminate, as far as possible, personal preferences and emotional responses. Institutions and organisations are increasingly suspicious of any process that delivers a different outcome from the one a computer would have delivered. We are urged to rely on procedure rather than experience, to follow rules rather than than making choices. No wonder computer programmes can imitate us so effectively.
You're right to mention Marvel movies. Sean Thomas, the British legacy media's most insistent enthusiast for AI, recently wrote a book and submitted it to an AI editor; he claimed that the editorial feedback he received from the machine was as good as any feedback he could hope to get from a real-life, flesh-and-blood editor. But the machine's advice was essentially to make the book more conventional: "Lack of clarity can be frustrating for readers who want a definitive answer," so "while ambiguity can be effective in some cases, it’s important to provide enough clues and hints to allow readers to draw their own conclusions about what happened." If that's the kind of advice a professional editor would give, perhaps it explains why so much modern fiction is so unenterprising. Well, OK, Thomas was writing genre fiction. But even modern literary fiction often seems to be written to a formula, insistent on clarity, determined to explain itself, and deeply reluctant to tolerate the kinds of ambiguity that were the mainstay of the novel in the hands of, say, Henry James, and were common in his heirs up to the 1990s. Read, say, the early Angus Wilson (singling him out somewhat arbitrarily since I have read all of his 1950s novels in the last few years), and marvel at his refusal to advance neat and tidy explanations for human behaviour. The kind of demands he makes on the ordinary intelligent reader are, in their quiet way, quite breathtaking.
Of course, as you suggest, the question these days is how many people can spot the difference.
AI IMO writes better than maybe 75% of people – it’s just that these 75% self-evaluate and just don’t publish anything (I certainly don’t publish long-form essays here or anywhere, let alone fiction).
Even if it’s stylistically bland, inexpressive output, that matches or exceeds vast majorities of humans. Three years ago, it beat maybe a third of people. Before that, it could not functionally write coherently at all, which frankly is also “human level output” for large proportions of the world population who ever functionally learned to write at all, i.e. not just shopping lists.
If the proportion of people it can’t best in each field keeps – let’s say – roughly halving every 2.5 years and my assessment above is correct, AI will be a 98.5th percentile writer in a decade. Still not a great novelist, but what percentage of humans is.
Progress is of course uneven in different fields – good writing is probably just not enough of a market to invest scarce and expensive ML talent in, plus you get a lot of good, angry prose written about you if you try for obvious reasons. E.g. in programming, not an artistic but arguably a creative field, the best current models are 90th+ percentile contributors only given natural language instruction.