AI food images will deceive you.
Photo by ChatGPT via Reddit
Artificial intelligence (AI) is revolutionizing our world, especially data-driven technology and marketing. But what about the culinary arts? Platforms like ChatGPT use data on ingredients, pairings, food chemistry, and past recipes to create, what it thinks, are solid flavor profiles. AI even creates innovative dishes, but when it comes to generating plausible recipes, that’s another deal.
The problem with AI recipes is that they lack human intuition. There is no brilliant chef behind the recipe or flavors, simply a robot full of algorithms. It doesn’t always understand the complexities of taste, mouthfeel, and aroma, nor does it account for cultural preferences… but it sure is entertaining.
AI Recipe Fails
ChatGPT suggests some strange ingredient pairings, like Thai green curry lasagna, garlic honey trifle, and a banana mushroom smoothie. One recipe for burnt lettuce soup calls for boiling lettuce for 30 minutes before broiling it at 500 degrees. Sadly, delicate vegetables tend to suffer when it comes time to AI-generated recipes since the algorithm doesn’t understand which ingredients should be cooked or how.
I’ve got to admit, some of the dish names AI comes up with are downright disgusting: Creamy Garlic Nightmare, Suspicious Onion Casserole, Spicy Butter Explosion, and — my personal favorite — the Bony Meat Cloud. Hungry yet?
AI also uses overly complex recipes, some calling for two kilos of butter. (I’m a classically trained chef, so I love and appreciate all the butter. But two kilos? That’s a bit much, even for Julia Child.) There are sandwiches with upwards of 50 ingredients. Conversely, ChatGPT suggests ingredients even Gordon Ramsay can’t get. Think moon-dried saffron petals, hand-foraged Himalayan ice basil, and organic meteor dust. (All fictional ingredients.)
If that’s not enough, the recipe for a deconstructed PB&J calls for making homemade brioche, roasting peanuts before grinding them into peanut butter, and cooking fresh grapes to reduce to a compote. There’s also an AI-generated pasta recipe that uses seven different pots, a 30-ingredient scrambled egg recipe, and a lasagna that cooks for 10 hours.
Utterly useless recipes include the infamous “crispy water” recipe and a methodology for preparing crock pot cold water. As the home chef, you must take precautions because AI also generates dangerous recipes like gasoline flambéed steak, bleach-marinated chicken, deep-fried ice cubes, and dishwasher salmon. (Don’t try those at home.)
Don’t Believe the Pics
This AI-generated pizza recipe called for strange ingredients.
Photo by Morgan Banno
AI-generated food pictures are everywhere. But don’t let your eyes deceive you. In the spirit of experimentation, I recreated a pizza that was an AI-generated image found on social media. While it looks pretty at first glance, the ingredients make no sense: raspberries purée, fresh basil, blackberries and…. tortellini?
So why does AI sometimes fail so miserably at recipe generation? It lacks common sense and culinary experience. It follows algorithms, not tastebuds. Often times, the results that are generated are at best, incomplete, and at worst, faulty data. Human creativity and intuition are paramount in recipe development, and it’s the one major flaws in using AI to generate recipes. AI can assist in meal planning and macros, but actual human taste testing is irreplaceable.
Your donations are tax-deductible to the full extent of the law. The Florida Community News Fund is a program administered by the Florida Press Foundation, tax ID #59-2449377, a 501 (C) (3) organization. All donations made at this link go to The Gabber Newspaper through the Florida Press Association, a 501(c)(3).
Please support local news and The Gabber Newspaper!