Last year, as an experiment, I baked cakes based on recipes generated by ChatGPT and Bard. It’s better than I expected – as much of each recipe doesn’t seem to have been created by artificial intelligence at all.
The recipes I was given looked eerily familiar to recipes I’d seen on food blogs or on Instagram. They were slightly altered, with only Bud (now called Gemini) bothering to provide attribution, linking to the recipe for Sally’s baking addiction. In some ways, this is not a new problem.I’ve seen too many plagiarisms No-Artificial intelligence recipes are circulating on the Internet. However, artificial intelligence makes the ethics of recipe copying even more fraught. Its sheer scale has the potential to create a world where human recipe developers are squeezed out by semi-random competition generated by artificial intelligence, but despite signs that this world is coming, we’re not yet in it.
Even without artificial intelligence involved, recipe similarity is a very sensitive topic in the food world. Some influencers can find themselves in the middle of a firestorm if another food blogger or cookbook author claims they stole their recipe. New York Times Several cases of alleged recipe plagiarism have been reported, including the 2021 case in which author Sharon Wee accused British chef Elizabeth Haigh of plagiarizing passages and recipes from her cookbooks. Then came the online hatred for popular food blogger Half Baked Harvest, who was repeatedly accused of plagiarizing two other popular food bloggers.
But food creators have little legal recourse if they think someone has stolen their work. Simple ingredient lists and instructions are considered non-copyrightable ideas. Many recipes have ingredients from oral traditions. Many are passed down through family members.Although some contain lengthy personal prefaces yes Despite copyrights, the biggest pushback against “stealing” recipes tends to be the daily satirical posts on Foodsnark Reddit, and the recipe world is largely governed by etiquette rather than law.
Like musical chords, you can combine a limited number of ingredients to get a recipe that’s not bad. Take a piece of pie crust. How many permutations of sugar, cream, and flour can one write before someone claims they’re just repeating the same recipe?
Large language models, such as those powering ChatGPT and Gemini, can take these permutations, parse them faster than humans, and quickly come up with a reasonably reliable recipe. Therefore, finding recipes that match specific dietary habits is often considered a good potential use for chatbots. On the other hand, let’s be clear, AI tools can’t actually prepare or eat food. They don’t really “know” whether a recipe works, just whether it fits the pattern of valid recipes.
Abi Balingit, a cookbook author and recipe developer who runs the blog The Dusky Kitchen, said she doesn’t create her recipes with artificial intelligence in mind. But she worries it could affect where food writers and developers can showcase their work.
“There is a gradient between good and bad things, and artificial intelligence will not make recipe development worse because there is no guarantee that what it comes out with will work well,” Balingit said. “But the nature of media is ephemeral and volatile, so I worry that publications may at some point turn to AI instead of recipe developers or chefs.”
Generative AI can still occasionally hallucinate and concoct things that are physically impossible, as many companies found out the hard way. Grocery delivery platform Instacart has partnered with OpenAI, which runs ChatGPT, to provide recipe images. Results include hot dogs with tomatoes inside and salmon Caesar salads that somehow create a lemon-lettuce mixture.The proportions are wrong – because Washington post Point out that the steak sizes in Instacart recipes can easily feed more people than planned. BuzzFeed It also launched an artificial intelligence tool to recommend recipes for its Tasty brand.
Balingit added that when people read someone’s recipes or watch them cook, they develop a level of trust. It’s the expertise of making food and actually tasting it.
This explains why I immediately felt the need to double-check the chatbot’s recipe. AI models can still hallucinate and grossly misjudge how ingredient quantities affect taste. For example, Google’s chatbot inexplicably doubled the eggs, which made the cake moist but also dense and gooey in a way that I didn’t like.
According to Balingit, one of the great things about being a recipe creator is the human connection.her recipes Mayuki Filled with imaginative dishes, she tells the story of her parents’ immigration to the United States from the Philippines, her childhood in California, and her current life in New York. She said AI has no cultural or nostalgic connection to food and eating, which historically are personal things that people share with others.
I feel the same way. Even though I never really cared who first invented the idea of cooking chicken in ginger broth, it remains my favorite Filipino dish, tinola. When I first started learning how to cook (thanks in large part to my dad kicking me out of the kitchen because I hated it), I searched the internet for “authentic” recipes. I chose to focus on the work of people who had a strong personal connection to the dish, which made me want to make the food they described so passionately.
ChatGPT and Bard can generate functional recipes. I know this because I follow them. But as someone who bakes these cakes, I know it’s cool and ordinary. My editor Adi Robertson compared one to boxed cake mix, while the other reminded me of a sad cafeteria cake. Of course, it’s just right. Yes, this is chocolate cake. But cake can mean so much more.
