Cheesy, Gooey, Melt-in-Your-Mouth A.I. Slop ⇥ defector.com
The latest craze in Facebook generative AI spam: cooking pages with tons of vaguely surreal artificially generated images of food.
I am not sure about “latest”, but this genre of A.I. slop seems to be increasing in volume to the extent even I have noticed on the open web. I find it repulsive for three reasons: first, developing recipes is difficult and, like many of the other creative practices mimicked by A.I., it deserves more credit than it gets; second, the results may be dangerous and nobody is incentivized to provide quality control; and, third, it is a depressing mimic of a fundamentally human practice.
Charlotte Engrav, reporting for NPR in September:
For years, chefs on YouTube and TikTok have staged cook-offs between “real” and AI recipes — where the “real” chefs often prevail. In 2022, Tasty compared a chocolate cake recipe generated by GPT-3 with one developed by a professional food writer. While the AI recipe baked up fine, the food writer’s recipe won in a blind taste test. The tasters preferred the food writer’s cake because it had a more nuanced, not-too-sweet flavor profile and a denser, moister crumb compared to the AI cake, which was sweeter and drier.
AI recipes can be dangerous too. Last year, Forbes reported that one AI recipe generator produced a recipe for “aromatic water mix” when a Twitter user prompted it to make a recipe with water, bleach and ammonia. The recipe actually produced deadly chlorine gas.
This article is not all eye-popping A.I. errors and trendy YouTube formats. It also contains a nuanced discussion about the ways different cooks and recipe developers are using A.I. services.
Katie Notopoulos, Business Insider:
I looked into a few Facebook pages that are posting what appear to be AI-generated recipes with AI-generated images. (How’d I come to suspect? The images had telltale signs of AI, like disappearing tines on a fork or weirdly shaped fingers or distorted edges.)
What I found most surprising: People are actually cooking these AI-generated recipes. Sometimes they’re even enjoying the results.
So I had to get in on the kitchen action myself. I made one of the salmon dishes — let’s call it “SalmonGPT.”
Ali Domrongchai, Defector:
However real the food and recipes might incidentally be comes second to the mechanics of getting clicks, likes, and shares. Creators aren’t taking a moment to read or check any of this work, the way anyone would when testing out a recipe. The goal is simple: mass-generating content to populate sham blogs as fast as possible. These AI-generated food pages aren’t just silly experiments, they’re businesses — like the rest of food media. By generating high-click content and using algorithms to mimic the viral successes of real food blogs, the accounts accumulate display-ad revenue and further clutter search results with garbage. It’s a machine designed solely to grab attention and profit; food was never the point.
This article is my primary link for a reason: if you read nothing else in full from this collection, you should at least read this one.
I have so far been underwhelmed by the small amount of experimentation I have done with ChatGPT. I am an enthusiastic home cook and have tried to get it to help me make menus for a week’s worth of meals, and for individual dinner parties. It loves suggesting skewers, for some reason, and out-of-season produce even when I specify otherwise. I can see how it may be plausible for basic recipes being more reliably generated, as there are many types of cooking based around a set of similar techniques but with individual ingredient substitutions.
I keep stumbling across recipes in the wild clearly generated by robots. Thanks to Google’s power over the web, they follow the same format as any other recipe blog, complete with a rambling introduction and history of every ingredient in the dish. Remember: traffic does not originate solely from Facebook, and demonstrating topical “expertise” is essential for search optimization purposes. The images are incredibly shiny and impossible to miss on Pinterest. That is how I found three A.I. recipe websites within just a few minutes, two of which — That Oven Feelin’ and Easy Family Recipes — have nearly a million followers each on Facebook. The latter is part of a network of similar sites that, if their hosting information is anything to go by, are operated by the same people. All of these websites have a lot of ads.
If the way you find recipes is by searching the web or Pinterest, I would bet you have stumbled across a generated recipe at least once in the past couple of years.
Craig Silverman, ProPublica:
Meta CEO Mark Zuckerberg also said in January that the company was removing or dialing back automated systems that reduce the spread of false information. At the same time, Meta is revamping a program that has paid bonuses to creators for content based on views and engagement, potentially pouring accelerant on the kind of false posts it once policed. The new Facebook Content Monetization program is currently invite-only, but Meta plans to make it widely available this year.
The upshot: a likely resurgence of incendiary false stories on Facebook, some of them funded by Meta, according to former professional Facebook hoaxsters and a former Meta data scientist who worked on trust and safety.
This is going to incentivize more slop of all kinds thanks in part to the business models of Google and Meta. Both companies have ways of rewarding people for being popular — Google Ads and Meta’s “Content Monetization” program. Even if you do not use Facebook, the open web you use is being flooded with slop like this. It offers no advantage over real recipe blogs — the exhausting narratives are still there — and it comes with a whiff of fraud.