Thanksgiving is all about family recipes with personal flair. Sure, lots and lots of families eat turkey and stuffing on the fourth Thursday in November, but only a few get your grandma’s special stuffing or my dad’s meticulous slow-roasted turkey. So how can AI compete?
The New York Times decided to find out. The paper’s food team turned to GPT-3, advanced technology created by Open AI which uses algorithms to generate text. Food reporter Priya Krishna started by feeding the AI tool personal details about her background and eating habits.
“I am originally from Texas and grew up in an Indian-American household,” Krishna wrote. “I like spicy flavours, Italian and Thai food and desserts that are not too sweet. Some of the ingredients I cook with a lot are chaat masala, miso, soy sauce, herbs and tomato paste.”
She then asked GPT-3 for a Thanksgiving menu. She followed 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’s not too sweet and a little spicy.
The result was an ambitious-sounding Thanksgiving menu: pumpkin spice chaat, green beans with miso and sesame seeds, side stuffing, roasted turkey with a soy ginger glaze, cranberry sauce that’s “not too sweet and a little spicy,” and pumpkin spice cake with orange cream cheese glaze. GPT-3 generated recipes for every dish, and the Time team also used DALL-EOpenAI’s image generation tool, to create visuals for each asset.
AI has generated recipes before. 2016, Janelle Shanea researcher who runs a blog named AI weirdness, used AI tools to create recipes and then blogged about them. At the time, her results were bizarre: They included recipes for “cream cheese soup,” “salmon-style chicken base,” and “chocolate pickle sauce,” wrote BuzzFeedby Andy Golder in 2017.
The ingredients include meaningless items like “husked rice” and “chopped flour,” she says Time. But the technology has come a long way in the past six years.
“What it does really well sounds plausible,” says Shane. “So if you weren’t paying attention and someone just read this recipe aloud to you, you’d say, ‘Oh yeah, that sounds like a very common recipe.'”
With recipes in hand, Krishna and her colleagues got to work cooking and tasting. How did the dishes turn out? In the words of Time food columnist Melissa Clark: “We’re not out of work.”
“The cake was firm and more savory 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 garlic clove to season a 12-pound bird, and no butter or oil; the result was dry and tasteless.”
For those looking to put AI to the test in their own kitchen, the Time published the recipes online.
While AI may not be replacing someone’s grandmother’s recipes anytime soon, it still has food-related potential. For example, researchers at the University of Illinois published one study earlier this year explored how machine learning models can help reduce food insecurity. FortuneDanielle Bernabe explains, “In certain cases, AI and machine learning enable organizations to quickly collect and interpret large amounts of data to evaluate areas of need: predicting where and why hunger arises and efficient food distribution.”