Sarah and Kaitlin Leung develop recipes with their parents for their blog, The works of life.
Christine Han
hide legend
toggle caption
Christine Han
Apple released its iOS 18 this month. The update, which comes alongside the latest iterations of the iPhone, Apple Watch, and AirPods, includes expanded AI applications, called “Apple Intelligence.” Apple isn’t the only company integrating AI into its operating system. Samsung’s S24 devices and the UI 6.1 update included elements of Galaxy AI support, and Google phones will soon feature Gemini AI as well.
Many companies have announced a feature on their new phones that lets users use AI as a recipe-generating assistant. In Apple Intelligence’s demo, a user asks Siri for a dinner menu with the ingredients they have, and the AI returns a list of recipes using those ingredients. While this sounds handy, most of the press about AI’s relationship with cooking has so far been negative.
Apple integrates ChatGPT access into its brand new operating system.
Apple
hide legend
toggle caption
Apple
For years, chefs on YouTube and TikTok have hosted cooking competitions between “real” recipes and AI recipes, with the “real” chefs often winning. In 2022, Tasty compared a GPT-3-generated chocolate cake recipe with one developed by a professional food writer. While the AI recipe baked well, the food writer’s recipe won in a blind taste test. Tasters preferred the food writer’s cake because it had a more nuanced flavor profile, was not too sweet, and had a denser, moister crumb compared to the AI’s cake, which was sweeter and drier.
AI recipes can also be dangerous. Last year, Forbes An AI recipe generator has reportedly produced a recipe for “aromatic water mix” when a Twitter user asked it to make a recipe with water, bleach, and ammonia. The recipe actually produced deadly chlorine gas.
With AI-generated recipes, home cooks risk being left with a failed meal or a life-threatening situation. For food bloggers and recipe creators, this technology could threaten their livelihoods.
Sarah and Kaitlin Leung are sisters who make up half of the family behind The works of lifea food blog focused on sharing “recipes, culinary exploits and travel.” They launched the blog in 2013 with their parents, Bill and Judy.
Recipes for The works of life Sarah starts with what’s called the “ideation phase.” “Sometimes we have a group conversation,” she says. “Sometimes it’s responding to recipe requests from readers. Sometimes it’s completely new and requires a lot of research and experimentation, going to a restaurant to eat that dish, watching videos, or scouring the Chinese internet for ideas.”
Once an idea is conceived, the Leungs test a recipe up to 40 times. “It took my dad about a year to come up with some of his recipes,” Sarah says. All four family members must approve each recipe before it’s published. “We know our readers trust us with their ingredients and their time, so we try to make sure our recipe is not only effective, but also easy to read and follow,” Sarah continues.
The process of developing this recipe is also about connection and cultural understanding for the sisters. “We realized we didn’t really know how to cook Chinese food,” Kaitlin said. “That really comes through in the blog. We’re still learning and always trying to make sure we’re finding new techniques and new ingredients.”
“The stories that surround these recipes and the connections we make with people through them are deeply human,” Sarah says. That’s why the sisters are skeptical of AI-generated recipes. “The machine doesn’t eat and it can’t taste. So what is it?”
Andrew Olson believes that AI has a place in the recipe development field. He is a software engineer who develops recipes for his food blog, One Ingredient Chef, which offers recipes featuring a whole, unprocessed ingredient.
In 2019, Olson began experimenting with GPT-2, a rudimentary version of ChatGPT software. “I was already thinking about how it could be used to develop recipes and help people find new and creative ways to cook,” Olson said.
Olson’s DishGen can generate recipes as well as photos of what the finished product might look like.
DishGen
hide legend
toggle caption
DishGen
In 2023, it launched DishGen, a tool that leverages AI for kitchen-specific results. On the website, users can enter a list of ingredients to generate a recipe that looks like it’s in a cookbook. Each recipe even includes a header note with a sensory description of the final product and suggestions for when and where to serve each dish. There are little touches throughout the recipe that hint at the recipe’s copycat style. The cheese is sprinkled “generously,” the textures are “harmonious,” and the muffins are “healthy.” Premium versions of the software even generate images of what the recipe’s final product might look like.
Olson is aware of the bad press. “Google tells people to put gasoline in their pasta,” he says. “So DishGen has put a lot of emphasis on safety.” If you provide ingredients that can have toxic combinations, such as chlorine gas components, the website won’t generate a recipe, but will return a short error message.
The Leungs don’t believe AI recipe generators can replicate sensory experiences and take into account the same variations and special touches that human recipe developers can. “What meat mix do you use? What seasonings do you use to get the right amount of meat? How much salt? Is the salt affected by adding cheese, which is salty?” Since AI doesn’t eat or taste the food, it instead amalgamates content scraped from the internet and uses pre-existing, human-tested recipes to inform the recipes it generates itself.
“These companies are taking content created by real people, without giving credit or compensation to the people who created that content to train their AI models, and then going into direct competition with those people who created that content. So it’s a huge existential threat,” Sarah says.
Olson sees it differently. “A lot of recipe development is taking inspiration from other recipes you’ve seen. Like, ‘Oh, that’s cool, but I could do it differently,’ or ‘I could add something else.’ I don’t see this technology as any different,” he says. “They’re taking inspiration from what’s publicly available, but they’re not plagiarizing it or reproducing it word for word.”
“I’m not totally pessimistic,” Sarah says. “AI – I think it can be used in a brainstorming context. You could talk about storage, how long a condiment can be kept in the fridge, or you could talk about a particular ingredient and explain it.”
Olson agrees. “I think food bloggers could use AI to be more creative and come up with new ideas,” he says, “but I don’t think the technology is mature enough to have a blog that’s entirely AI-generated, although that would be an interesting concept. Maybe someone should try it and see how it goes.”
As the Leungs prepare for AI technology to reach this point, they’re making sure their blog won’t be mistaken for an AI-generated product by drawing on their family stories. Many home cooks have long complained about the long and sometimes unrelated stories they have to wade through to find a recipe in a blog post. “Weirdly,” Sarah notes, “I think people are going to look for these markers that someone created. Like it’s a story.”
Suzanne Nuyen I edited this story.
