Something for the weekend - AI slop is a thing, but it's sloppy thinking to use it as a label for everything you don't approve of

> We have a word for AI content. It's called slop - and people don't like it. My eye was caught by that assertion from Steven Huffman, co-founder and CEO of Reddit this week as it reflects an underlying mindset in society that is driving me mad. When in doubt, dismiss anything you don’t like as ‘AI slop’. In business terms, it’s a phrase legitimately used to sum up mass-produced articles and content written by an LLM, not a journalist or an analyst. (Basically, think the polar opposite of anything you’ll ever find on diginomica and work back from there!). But it’s a great catch-all dismissive label. You don’t like the latest director’s cut of a classic sci-fi movie? AI slop! A Doctor Who fan experiments with using AI to produce amateur adventures on YouTube? AI slop! Someone you don’t rate releases an annoyingly catchy pop song that makes millions? AI-slop! It’s a phrase that’s come into the vernacular remarkably quickly. It seems not that long ago that if a Princess of Wales doctored some family photos before release to the media, this was excused as an over-eager case of experimenting with Photoshop. Now it would, without doubt, be ‘Kate’s AI slop!’. And so it goes on. Ad nauseam. And like all over-used terms, it loses its impact and meaning over time. Slop all around Now, there is slop out there and lots and lots of it. I’m not even going to begin to try to counter that reality. But my problem is how quickly it has become a default, go-to comment to slap on anything that meets with anyone’s disapproval. Some ‘AI slop’ offends, some amuses, some does both depending on your perspective. Donald Trump as the Pope, shared online shortly before the real Pope’s funeral last year. Doctored scenes from a hurricane disaster zone. Dolly Parton being welcomed into heaven by other dead music industry legends. Which offends? Which disgusts? Which amuses? Of course, none of this is new. Nearly 30 years ago the tacky amongst us could buy a commemorative t-shirt or poster with Princess Diana walking hand-in-hand with Mother Theresa through the pearly gates. That wasn’t ‘AI slop’, it was just tasteless opportunism. And yes, I do have a mug somewhere which has the late Queen Elizabeth also walking into the sunset hand-in-hand with Paddington Bear. (Paddington as the Grim Reaper's ursine sidekick walking off with the latest dead celebrity is still a thing today!) So what are we defining as ‘AI slop’? Where does the slop end and something become, for example, satire. Anyone who follows The Crewthorne Gazette might well give you an argument here. Those who are the subjects of its AI-powered mockery would have a different view, no doubt... Show me your dinner! Back to Reddit’s Huffman to see where slop begins and ends from his perspective. He points to the fastest-growing community this year on Reddit, something called GirlDinnerDiaries: > A post on GirlDinnerDiaries is a picture of food, it's like microwaved food, and then a long story about their bad boyfriend. It went from zero to 6 million readers this year, and it really exploded this summer. If you see it in the feed, it looks like a picture of chicken nuggets. > > > There's many, many of these every day and there's thousands of comments on every one of these posts from this community. And it's not just women. I mean I know about it because I'm reading it too. And so the opportunity in the feed is to create more of these magical moments for our users, because this content doesn't exist anywhere else online and it's really, really interesting. You what? Oh, come on, people have been photographing their dinner since Instagram was born. It’s a stupid, annoying habit that fully-grown adults ought to have gotten over by now, but it’s hardly unique and original content, is it? But for Huffman: > It's a story that's literally never been told before. And it’s not ‘AI slop’, it seems: > The example I just gave about GirlDinnerDiaries, AI can't do that, right? AI is not a bad boyfriend yet. Maybe we'll get there. But then I'm sure there'll be a subreddit for like whatever, my AI boyfriend is so supportive, or is too supportive, or whatever it is. Now, personally I would have seen AI as something of a threat to the likes of Reddit, but Hoffman argues: > My point is, there's all of these experiences and perspectives that come from other people, and the AI world creates a premium on these human experiences. He goes on to dismiss what he calls the mis-conception that AI kills Reddit: > The principle here is that in an AI world, there's a premium on human experiences. All of the things I've mentioned about Reddit, like all of the human perspective that drives those purchasing decisions, all of those stories that people tell on Reddit but don't tell anywhere else, this is like the GirlDinnerDiaries example...AI can't do that. So AI can replace a lot of commodity content...but it's not going to hold a conversation about the things you care about. And he has another bold claim to stake: > Ironically, AI doesn't exist without Reddit. Er...? Huffman plows on: > Reddit's data has been used to create every AI model that you've heard of, some with our permission and some without. And so the value of that data is tremendous. And I don't think anybody is debating this at this point. And furthermore, the companies aren't even denying having used Reddit data at this point. It used to be maybe an open secret, now it's just open. But it is what it is, he acknowledges: > For us, I think whether these relationships are formalized or not, Reddit wins because we have the data. We have the content that the models want. We have the content that the Internet consumer wants. He concludes: > It's not dissimilar from the journey we went through over the last 10 years where Reddit stands out because it's not social media. Now we're in the era of Reddit is not AI. Now the paradox there is we use AI all over our product, we can make a better product, a safer product. Reddit's content is used to create AI, but ironically, Reddit is for humans. Human nature And lots and lots of humans do like ‘AI slop’. And lots of humans don’t. That’s fine. What matters is surely our ability to remain flexible and alert to the dangers it poses. If we sit there gobbling up AI-generated content as though it’s gospel truth or another picture of the delicious Chicken Cesar Salad I had for dinner last night then we are in danger of our critical faculties being eroded to a point that is societally suicidal. As someone who is tech savvy, I already find myself on a daily basis doing double-takes when presented with images or videos and wondering if it is AI-doctored or not. I really don’t like the sense of permanent paranoia that this has instilled in me. I’m reasonably confident that human beings usually only have four fingers and a thumb, so when I’m shown an image of a celebrity with six digits on one hand, I’m a wee bit doubtful of the provenance. Silly old celebs, touching up their appearance. But then it was ever thus, from the days of vaseline on the camera lens and face-lifting sticky tape behind the ears. But when President Volodymyr Zelenskyy appears on video urging Ukrainian troops to lay down their arms to the Russian invaders, that’s a whole new scale of dangerous. Or when US voters get a phone call from ‘Joe Biden’ urging them not to vote in the then upcoming elections. That’s clearly ‘AI slop’ sneer some commentators, but where does slop end and disinformation and propaganda begin. And how far can we rely on the critical faculties of the 24/7 news channel population to be able to make the distinction. In George Orwell’s 1984, “The Party told you to reject the evidence of your eyes and ears. It was their final, most essential command.” Today there are many siren voices around us who earnestly want us to accept the ‘evidence’ of our eyes and ears as the most essential command. So, let’s lose the all-encompassing ‘slop’ label and start to treat this as a more nuanced