There are ten things that people commonly fight about on Thanksgiving. The top five are coming out of the closet, accusing someone of being an alcoholic, issues with finances, getting divorced, married, or pregnant, and the fact that you are simply not hungry.
At the bottom of this list are issues such as asking questions about when you will eat and if you may bring your partner to dinner, personal failures, calling the chef a bad cook, and taboo subjects such as politics, religion, and sex. A new point of contention has recently been added to that list as well, however—namely, if it is okay to allow artificial intelligence to create recipes for the Thanksgiving meal.
Technology has revolutionized the way we work, how we spend our leisure time, and the manner in which we communicate with each other. Now, it might also be able to generate the type of food we eat thanks to innovations in the last decade.
The New York Times generated the scheme, having requested a Thanksgiving menu from an artificial intelligence system.
It has already been proven that robots can make some foods faster than humans.
Becky Hughes, The New York Times Cooking social media editor, took that concept one step further, however, and decided to see if they could take over the Thanksgiving cooking. Priya Krishna, a food correspondent for The New York Times, was the one who then managed the experiment.
Krishna was intrigued by having a neural network come up with the whole holiday meal. She therefore asked a Generative Pre-trained Transformer called GPT-3 to write the corresponding recipes for that meal. Another program called DALL-E created the images.
According to the journalist, recipes included not only precise instructions but also ingredients, specific measurements, and personalized introductory notes.
In speaking to CBC News, Krishna, who was asked about the motivation behind this project said, “We thought, you know, artificial intelligence has gotten really smart. Do we think it can make a better…menu than us humans, who sort of spend year after year trying to [re-invent] the wheel on Thanksgiving? ”
Thanksgiving recipes and the GPT-3 system
Krishna’s reasons for questioning if their ability to do so was better were certain human limitations. Among these is, for instance, the impossibility of reading every single recipe on the internet for pumpkin pie or mashed potatoes. Yet, computers, as she pointed out in the article, actually have this ability.
“All of the recipes it’s generating are original,” she maintained. “It’s not just copying and pasting from the internet.”
GPT-3 “came up with a naan stuffing that had 32 different ingredients,” she said. “That was wild. A roast turkey with a soy glaze that only called for one clove of garlic for a 12-pound (5.4 kilo) bird. And then a really confusing one called pumpkin spice chaat, which was just pumpkin puree with spices and cilantro and lime.” Four cooking columnists for The New York Times then tried the recipes.
artificial intelligence can write poetry and play jeopardy…I thought, what if it can do Thanksgiving better than we can? I got to work with @CadeMetz on answering this very fascinating and creepy question. https://t.co/5agRhbkr1O
— Priya Krishna (@priyakrishna) November 4, 2022
So how did GPT-3 recipes fare? Well, in Krishna’s words, “It started out really hopeful. You know, first you sauté the onions in a pan…mix ingredients together for a cake batter. You sort of felt like you’re going through the correct motions that should lead you to a good end result. But then, ultimately,…most of the dishes did not turn out great.”
Nevertheless, she did add that the A.I. system shows potential.