You may have heard of artificial intelligence-generated poetry, comics, or award-winning art. How about an AI-generated Thanksgiving menu?
Thanksgiving is all about family recipes with a personal flair. Sure, tons of families eat turkey and stuffing on the fourth Thursday of November, but few get your grandmother’s special stuffing or my dad’s meticulously slow-roasted turkey. So how could AI compete?
That New York Times decided to find out. The newspaper’s food team turned to GPT-3, an advanced technology developed by OpenAI that uses algorithms to generate text. Food reporter Priya Krishna began feeding the AI tool with personal data about her background and eating habits.
“I’m originally from Texas and grew up in an Indian-American household,” Krishna wrote. “I love spicy flavors, Italian and Thai food, and desserts that aren’t too sweet. Some ingredients I cook with frequently are chaat masala, miso, soy sauce, herbs and tomato paste.”
Then she asked GPT-3 for a Thanksgiving meal. She followed up with specific requests: “Show me some desserts tailored to my taste preferences. Show me a non-traditional Thanksgiving recipe. Show me a recipe for cranberry sauce that isn’t too sweet and is a little spicy.”
The result was an ambitious-sounding Thanksgiving meal: pumpkin spice chaat, green beans with miso and sesame seeds, naan stuffing, roast turkey with soy-ginger glaze, cranberry sauce that’s “not too sweet and a little spicy,” and Pumpkin Spice Cake with Orange Cream Cheese Frosting. GPT-3 generated recipes for each dish and the Times The team also used DALL-E, OpenAI’s image generation tool, to create visualizations for each item.
AI has generated recipes before. In 2016, Janelle called Shane, a researcher who runs a blog AI madness, used AI tools to create recipes, then blogged about it. At the time, her findings were whimsical: they included recipes for “cream cheese soup,” “salmon beef style chicken crust,” and “chocolate pickle sauce,” she wrote BuzzFeed‘s Andy Golder in 2017.
The ingredients included meaningless things like “husked rice” and “chopped flour,” she says Times. But the technology has come a long way in the last six years.
“What makes it really good sounds plausible,” says Shane. “So if you weren’t paying attention and someone read this recipe out loud to you, you’d be like, ‘Oh yeah, that sounds like a very ordinary recipe.'”
With recipes in hand, Krishna and her colleagues set about cooking and tasting. How did the dishes turn out? In the words of Times Food columnist Melissa Clark: “We’re not unemployed.”
“The cake was dense and savory rather than sweet,” writes Krishna. “The naan filling tasted like a chana masala and a fruitcake that got into a bar fight. The roast turkey recipe called for a single clove of garlic to flavor a 12-pound bird and no butter or oil; the result was dry and tasteless.”
For those who want to put AI to the test in their own kitchen, the Times published the recipes online.
While AI won’t be replacing grandma’s recipes anytime soon, it still has potential in food. For example, earlier this year researchers at the University of Illinois published a study examining how machine learning models could help reduce food insecurity. wealthDanielle Bernabe of , explains: “In certain cases, AI and machine learning enable companies to quickly collect and interpret large amounts of data to assess areas of need: predicting where and why hunger occurs, and efficient food distribution.”