Why is Slop Bad?
This is like QVC for AI
A few weeks ago my girlfriend and I were at a hotel and naturally we spent hours watching weird hotel room television: a world full of ancient stories beyond culture, beyond entertainment, beyond everything. We landed on QVC and could not look away; there were women selling tacky overpriced face creams which came in strange bejewelled tubs. These women were tired, their spirits charred by the flaming bullshit that had to repeatedly exit their mouths, their faces blotted and puckered from demoing the face cream for several consecutive hours, their smiles forced, their voices horse. They were in hell. We realised after a while that they actually weren’t allowed to stop talking until they had sold every tub.
These women, selling promises of eternal youth and not face cream, were — or still are lol —trapped in the surplus dimension, and the only way out was/is to tell the public that the cream would change their lives forever. This is normal, this is sales. It’s not about the product and it never has been; the face cream may as well not exist — the active grift is the real product here. And this methodology is what keeps the tech industry going. The true innovators are not the ones who try and sell you iPhones; they are the ones who sell you Pure Triple Filtered Nothing, while explaining that they need $10bn to get started.
There’s something maniacal and gratuitous about the amount of AI spending that’s happening now. Investments in data centre development don’t seem to be slowing down, even though there is no real end product, use-case, or clear ROI here. Deutsche Bank have warned that this is untenable and the spending cannot continue. But also if it stops it will reveal an economy ravaged by circular deals, forged by the men who literally cannot stop talking until we’ve all accepted that AGI is real, and is definitely coming. This is like witnessing a pump and dump scheme where the dump part never ever arrives. It’s all pump! Now all we have to do is just wait for the bubble to burst so we can collect our reward of unparalleled economic turmoil, but I’m not sure there will be a huge punctuating event where everyone gives up and stops spending. Sarah Myers-West from AI Now recently characterised this as less of a traditional boom-and-bust and more of a new status quo where AI men and their dumb financiers refuse to see that they are not creating any value, and are just kidding themselves about how long they can keep this dream alive. But this isn’t a hole you can spend your way out of.
There are projections that suggest the industry will need $2 trillion in revenue to meet the demand that has been fabricated by AI companies. Last week Matt Levine reminded us that “the essence of finance is time travel” in his piece about how OpenAI keep making deals. Their latest one with Nvidia is very circular, and very criticised by analysts; meanwhile Oracle is financing their AI infrastructure buildout with billions of dollars of debt. Sure yes let’s just deal in an ever swirling Möbius strip of abstractions until we all drown in spongebob nazi AI slop.
But actually, if Sora 2 represents the future that we’re burning cash for in the present, then that’s fine. I will be happy if everything stops here and they all finally give up and move to their weird synthetic techno-libertarian islands with their robot girlfriends and human wives. But I don’t think this is going to happen, in which case we have to start thinking about a world beyond slop, or in harmony with slop. Cast your mind back to the dark ages when we didn’t even have ChatGPT. I call it 2019. GPT-2 came out and it was nothing like the sexy product drops of nowadays. Sam Altman didn’t even tweet about it. He was barely a public figure, he was just posting the usual inane shit about seed funding like every other startup founder. And the marketing team at OpenAI decided to hold back with a GPT-2 general release, over fears of what it might do if it fell into the wrong hands. But once ChatGPT came out they took the inverse approach, and said everyone could have as much generative AI as they wanted. ‘It’s up to you to find a use-case! Try not to fall in love!’ they said.
This all feeds in to the idea that AI is a shared project, and we all have to work on it together to make it ‘good’, even if that means you become slightly unwell. But so many of us can plainly see that the outputs AI is giving us are not good at all. That is why we call it slop; because it looks like complete shit and it seems to serve no purpose. The unfortunate thing here is that people can’t stop talking about just how bad the slop is. Commentators much more powerful than I keep saying that the internet is ‘flooded’ with slop now. First of all, you can’t flood the internet with anything. It’s way too big and fragmented. Secondly, is all slop bad? This week Charlie Warzel, who probably has his feeds set to max slop because he thinks he needs to do that to do his job, has been lamenting over older family members sharing crappy AI-generated reels of complete slop in group chats. Even though it’s dumb, can’t people just like what they like?
Slop is messy and incoherent. It oscillates between being impossible to ignore, and seemingly harmless. This is very much like modern conservative discourse — and it’s why far-right groups have been using it to their advantage very well. Eryk Salvaggio has written about how slop can be used to bait users into feeling outrage over whatever gripes The Right might have that week, even if those posting/sharing know that the images are completely fake. And I think about this a lot. However I also think about what Cory Doctorow has always said about how, historically, new media-makers would infuriate the incumbents, and be accused of piracy, or of dumbing things down, trivialising a craft, or just outright obliterating culture — but then the new media becomes the norm anyway, and cultures evolve to fit. And this cycle repeats. E.g. when Betamax came out (this was the precursor to VHS), Hollywood studios hated it and sued Sony. When Napster came out, Sony (who over the years had consolidated their media empire) hated that and so sued Napster.
It’s hard to place slop as just a new kind of media, because the people who like it, or just don’t care, are kids and old people. Probably because their tastes sit outside of what aging millennials consider to be high-quality media. Old people on Facebook really love that Jesus made of prawns and kids love Italian Brainrot; which is essentially a super-meme with its own extended universe of lore, and multiple online communities who keep the wiki up to date with characters, or create new ones. I’ve seen videos of kids talking about Italian Brainrot as if it’s a TV show, even though in the traditional sense it doesn’t resemble a TV show in any way: the videos are super short, mostly nonsense, and if there is any kind of narrative it is without dialogue, and without visual consistency. Because it’s all AI generated. There’s a piece of wood with a face and a thousand-yard stare. There’s a cup of cappuccino which is also a ballerina. There’s this really angry jacked orange. These characters form relationships and have arcs and then people talk about them online — this all follows the rules of fan fiction, but there was no external canonical source material to work from; its all come directly from the fandom itself. The fans decide what is canon.
I think there’s a lot going on here: there are valid concerns that fascists are using slop to progress their agenda. This is true, because a central part of fascism is using the media tools at your disposal to spread propaganda. Maybe the outraged old media left could try this just once. Or at least stop getting confused and infuriated that it’s working against them. Tied into their confusion is snobbery: although completely different, this moment does remind me a bit of 2015 when the intellectual elites couldn’t stop complaining about listicles. Yes, people are making weird dumb shit that you don’t understand — welcome to the internet. The discourse feels very bogged down in dreary one-dimensional takes insisting that all AI-generated media is always bad no matter what. If you only look at Meta’s cringe as fuck Vibes feed then yes of course it’s bad. But teenagers will always find a way to make bizarre claggy nonsense online together, in the most obscure places, just like we did when we were lonely unhinged little gremlins on IRC or whatever, looking for strangers to validate us lol.
So slop itself isn’t inherently bad; there’s nothing wrong with kids entertaining themselves with meaningless drivel. Didn’t you ever watch MTV Cribs when you were 17? That’s 2002 slop. The Actual Bad lies in the fact that a handful of tech companies non-consensually scraped the internet to train their slop machines, and are now betting our life savings on all of this being profitable (maybe they should start selling brainrot merch like everyone else lol). This movement has also skewed our conversations about art and creative industries. Tech bros are some of the most uncreative meat-headed incubators for human mediocrity on earth — and yet they are now piloting narratives around what art is. The visions they lay out refuse to account for a creative world that lives outside of the logics of expansion and profit: e.g. with a service like Elevenlabs, you can replicate Judy Garland’s voice (and other dead celebrities) and stay stuck in the past forever, for a price. When we ask ourselves what will happen to artists if we can suddenly create anything at the click of a button, we are asking the wrong question. This question assumes all art is digital. Artists have been making, and continue to make, the most unhinged conceptual nonsense you’ve ever seen without ever touching AI, or without AI ever touching them. AI doesn’t have to govern this.
Equally, artists have also been commercialising their work in a cutthroat, elitist, exploitative industry for generations. AI should also not govern this — because it will only make things worse. The tech industry wants you to think that digitisation means democratisation, but it doesn’t. By forcing everything to be ‘social’ (the Vibes and Sora feeds operate like vertical video TikTok clones, basically) the big tech firms continue to facilitate that pernicious collapse of the producer-consumer binary, and completely miscomprehend the world that they are trying to build. If you look at things like Showrunner (it’s like AI Netflix), the sell is that you can generate whatever you want to your exact specifications — as a complete individual. If you can generate your own entertainment, exactly as you want it, with no outside help, why bother sharing it or consuming anything from anyone else? In this frame, generated media is to television what cars are to transport: everyone gets their own one, and you don’t have to interact with others to get to where you need to go. It’s sad.
Which is why when I gaze into the brainrot cinematic universe, I actually feel hopeful. Communities are socialising around making weird content together without relying on one tool, one platform, or one way of doing anything. Meme culture transcends a company telling you what the future is. Brands famously co-opt meme culture to ride engagement waves because they aren’t allowed to be edgy. Do not fear slop, just try better to understand it. And remember that the AI men aren’t excited by creativity and they don’t care about any of this: they are just so deep in the hole, that they literally cannot stop talking until they’ve sold all their overpriced snake oil face cream.
This article comes at such a perfect time! I realy loved your line about 'Pure Triple Filtered Nothing'. It totally nails what's going on with so much AI spending right now. You described it so well. Seriously insightful!