Earlier this month I saw Vin (@vin on here) post a photo of a croissant shaped like a dinosaur, and suggested it looked AI-generated. I agreed - I explained why in my post about detecting AI fakes - and then didn't think much more of it. But the next day on Instagram, I saw exactly the same image by its original poster, liked hundreds of thousands of times. Normally I would've skated on right by it, but the one comment I saw on the post made me think twice, and led me down a fun if... mind-bending rabbit hole that I'm going to take you down now too, about AI fakes, social media, and brand management. Off we go.
Let's do it in the order I came across it. I opened the comments on the post because I was curious about whether anyone had clocked it as AI or not. On Twitter, for whatever reason but possibly because Twitter's algorithms prefer to get people to argue with each other, AI-generated fakes tend to have comments pointing this out fairly prominently underneath them. But this post didn't have any of that, people seemed pretty enthusiastic. The comment that really threw me, though, was this:
Up until this point I hadn't realised what the account posting the image actually was. I assumed it was one of those viral content mills that just put out reposts of other people's content, with a name like "Just Millennial Things" or "Our Beautiful World 😍" - I didn't even bother looking. But no, this was actually a restaurant, called Ethos, based in Austin, Texas. The photo of the croissant dinosaur was something they were actually claiming they had made, and was on sale. But it wasn't. I mean it couldn't be. Viral junk using AI I understood, but why was a restaurant promoting something it couldn't possibly offer?
Naturally, like any reasonable human being with a life to live and a joyful world outside their window to enjoy, I went and scrolled through their feed.
The first few posts I saw were pretty normal, although all definitely AI-generated. They followed a similar pattern to the croissantasaurus: they were unlikely twists on existing food or ideas, usually with something intentionally weird about them. A block of soap that is made of coriander (because some people experience the taste as soapy); a Philly cheesecake topped with Wagyu beef; "Italian Nachos" made with mozarella and Italian sausage. The croissant had gone viral because it has something these all lacked: it was immediately, visually punchy in a way none of these really are. But knowing it was generated by AI, it was easy now to see that all of these were too.
In fact, not just the image, but the concept, and the post text. Unlike my post on deepfakes, I really can't tell you exactly why I think this is, I can't conclusively prove it, just that I've read a lot of AI-generated text, and the cadence, tone, choice of words, even the way it blends ideas in a very inoffensive and surface-level way - there's no doubt in my mind that every part of these posts, from the original concept to the text and images, were all generated with AI tools. So at this point, safe to say, I'm now pretty curious.
As I scroll down though, the posts get weirder. Bearing in mind I'm now assuming every single post is fake, but now I'm coming across posts that feature... people. And not in a vague way, they feature people in a really, really unnecessarily specific and detailed way. For example, after Valentine's Day they posted about two of their staff members falling in love and getting married at the restaurant. Except the images are fake. And in a rare slip-up, they're so fake this time that people actually notice. AI image generators aren't really designed for consistency, and some commenters note that the happy couple don't actually look like the same people in all the images.
This post is the last time they post an image of people - I don't know if it's because they felt it was too hard to get it right, or if the products were just going more viral. They have lots of posts about people before then though, but more generic and crowd-based. They're also madly, wildly, stupidly unhinged. This is my favourite, a post at Christmas about how they are going to be open over the holidays, that their staff aren't happy about it, with an attached picture of their unhappy staff in santa hats. At this point I google "signs you may be experiencing aphasia" just to double check I'm still on the same perceptual plane of existence I was an hour ago.
There's posts about how they won fake awards (these ones aren't even AI generated, they just... declared they won some awards for some reason); posts about their exciting brand collaboration with LEGO that definitely does not exist; posts about how they're sold out for the month of January, posts about... GILF night at Hooters. I'm so confused at this point. Almost all of their comments are enthusiastic, asking how to order, asking for recipes. Some are playing along (maybe?) or just goofing on them, pretending they've attended and praising the restaurant for its great food and service. It's bizarre.
I figure at this point I'm not going to ever understand what is going on here, and my plan was to just write this post, share what I've shared so far, and laugh with you at it. But then, a week later, I came back to show my friends and I noticed a change. Since the croissant post went viral - more than twenty times anything their posts had done before - they had made one huge change. They'd registered a website.
I don't have the space to go over everything that is bonkers about this website, but I can tell you a few things: first, obviously, they have a merch store. Like anyone else who has gone viral, they figure maybe they can convince a few people to pay £35 for a t-shirt with a croissant dinosaur on it. Maybe they can, I dread to find out.
They also have some articles on their website! These articles have definitely, definitely been written by an AI. Incredibly, in what is absolutely a joke on the part of the person running this, they are about authenticity in marketing food.
They have a reservations page too, which explains they only take reservations at the start of the month - such an exclusive place! They also list their staff, which again, props to whoever wrote this because this is the funniest part of this entire parody:
Heath Elderberry, Senior Forager? GIUSEPPE FUSILLI!? These are so wild I almost considered if they were just human-written gags but they're definitely AI-generated. Good lord. Anyway the website just makes it a wilder story to me, and it's obvious the restaurant doesn't exist, but at least the merch store sort of helps me understand why someone would try to do this. Kinda. I figure that's definitely the end of the road now. But there is one thing I haven't done yet, for obvious reasons, which is: try to make a reservation. It's literally the only thing left on the website that's interactive. I mean why would you click on it, the restaurant doesn't exist, what would the button even do.
Click here to find out (safe for work, although... weird).
It links to a website called Eel Slap, which displays an interactive gif of a man getting slapped (one would assume, by an eel). It's the final admission that this whole thing is a joke, if you dig far enough to try and actually reserve a space at this nonexistent restaurant. And at the bottom there's a tiny 'made by' link that links to...
Act Normal, a... I don't even know what you'd call them. Marketing company? Brand manager? Who knows. It's safe to assume these are the owners of the Insta page, although the WHOIS database entries and links don't necessarily confirm it. Whether it's one of their 'just for fun' project or proof of concept for something else, I assume they're enjoying experimenting with AI. Part of me doesn't want to be too hard on them, another part of me is just deeply tired, exhausted even, at the extent to which the Internet has been absolutely immolated from the inside out over the last fifteen years, and I don't really know why you'd want to find new ways to accelerate that. But there you go.
The story ends here, so if you just wanted a weird story about a weird thing I found on the Internet, that's all you need. However, because I can't help myself, I have a little postscript about AI and creativity and the Internet that you can read below. There's no jokes though, and I use the word capitalism, so if you wanna just bounce I won't blame you. In the meantime, if you'd like to keep up with things I write and make, I have a little Discord server, and you can find my personal site here with research, games and more.
Sure, But She's A Real Fake
Is Ethos a success or not? I'm completely speculating on the point of the project, maybe someone was just goofing around roleplaying for fun, but I'm guessing that there's some bigger experiment here to show the impact of AI tools on marketing. This restaurant doesn't even exist and we were able to get 32,000 followers and posts with almost a million likes. Imagine what we can do for a company that does exist! I see a lot of people talking about how people are finished, there's so much that AI can do, pretty soon we'll all be able to promote our businesses with glossy, beautiful, highly viral posts. Someone told me that generative AI is forecast to add $7tn - that's trillion - dollars to the global economy. What does that mean? It means absolutely nothing. But he said it, and the room of people he said it to believed him.
However, much like the post where I talked about the benefits of AI for offices and programmers, this approach to deploying generative AI collapses as soon as it becomes normalised or popular. The reason Ethos was able to attract attention is because people use Instagram under the assumption that they are finding real things; because their posts were compared against a majority of people who are unable to create images of physically impossible croissants; and because they never had to follow through on a single one of their ideas, or worry about disappointing their customers or being found out. The contrast and the breaking of rules is what enables it to work. It's simply not universalisable for all of us to get 980,000 likes on posts promoting our work. People can only do that if they win the competition the social media platform is running. The social media site has to pretend that you win by making the best stuff. In reality, there are other ways you can win, by breaking the rules (like lying about things you've made) or spending a lot of money (accessing production equipment, tools, software that others can't).
If we all were generating marketing with AI, the baseline would rise to the point where none of this would be interesting or meaningful any more. Photorealism would be banal and dull. And you'd still be overlooked in favour of large companies who can spend more on their content marketing, or by people making fake content to go viral. This is a zero sum game and there is no computer program in the world that can fix the odds for you while capitalism is the underlying structure. AI tools can impact the world positively in some very specific ways - but they aren't going to equalise systems that are rigged. Pretending that they are, and encouraging others to endorse them, or even just giving up and declaring that we've got no choice, all of these routes lead to us endorsing companies and products that are actively harming our world. A tiny branding company in northern europe playing silly games on Instagram isn't doing that much damage - but it's emblematic of where this technology is being leveraged, the thinking behind it, and the benefits people believe it has.
Earlier this week I went to hear Amazon sales managers talk about generative AI in order to try and sell it to university staff. What I saw was an incredibly weak sales pitch, even weaker than I thought it would be - I might write a little about it soon if I feel it won't breach anything. But one thing that did shine through, again and again, is that companies selling this technology are desperate to convince you it is inevitable and that the people who get their first will benefit the most. In that respect it's no different to the sales pitches you heard for bitcoin, for web3, for VR, for any other tech fad.
As funny as this story was, it makes me sad to watch the Internet slowly become non-functioning. We exaggerate its effect sometimes - for years people have called other human beings bots just because they didn't understand how they post, for example. But the core of the Internet grew smaller and denser over the 2010s, like urbanisation drawing rural communities into tight, dense metropolises, and these dense online cities are crumbling. If Cohost dies tomorrow, I know I'll find new places on the Internet to exist and breathe and create. But for many people in the world, and certainly the generations coming after us, we are leaving nothing for them to build on, nowhere to learn and grow in safety and freedom, no history to cherish or laugh at. Just an infinite scroll of fake meals that cannot be tasted, and will never be cooked.
