Photo: W. Scott Mitchell Photography
Chef Ian Ramirez, founder of Mad Honey Culinary Studio, known for his culinary creativity, first as a university chef and more recently at his own company, has spent the last few years experimenting with how artificial intelligence (AI) can augment human intelligence (HI) in tasks such as recipes and menu cycles.
“A lot of people have been making fun of AI over the last year,” said Ramirez, speaking at the Foodservice Directors Menu Direction conference, as he gave examples of the strange, mysterious fingers the AI has produced, as well as a response to a prompt: “Salmon jumping in a river.” The prompt showed a perfectly cooked salmon fillet swimming merrily up a river. In response to another prompt: “How many rocks should I eat?”, the AI recommended, “You should eat one small rock a day.” And when the AI was asked for advice on how to stick cheese to pizza, the answer was “non-toxic glue.” In another example, the AI said that cooking spaghetti sauce with gasoline would make it even spicier. Wow.
Yes, silly things have happened, but jokes aside, experts in many fields are asking themselves how to coexist with an AI that gets smarter and learns every day. Many human experts are feeling a bit uneasy, and it's understandable.
“There's a lot of resistance from businesses to adopting AI, and there's a reason for that,” Ramirez says. “Data security is one reason, so companies may not want to use AI. Another big reason is they don't know if what they're using may be someone else's intellectual property. There are so many unknowns right now, and that's why American companies don't want to use AI.”
Ramirez said that as consumers, many of us use AI in some way without even realizing it, citing online ordering, optimized food delivery times, drive-thru technology, etc. “My wife works in marketing, and she gave me this example: You pick a lipstick color on Sephora, and another site suggests a dress to go with it,” he said.
For chefs and managers, AI is already helping with inventory management, procurement and streamlining operations through workflow tracking, for example by measuring the steps employees take in the kitchen throughout the day to identify wasteful steps.
“That's really interesting and really cool, but I don't know how intrusive it is and how comfortable we are with that,” Ramirez said. “But monitoring the kitchen is something we all do. How many steps are being taken, how do we shorten this, how do we make this easier for people?”
He noted that another way to use AI is to insert screenshots of complex mathematical formulas or graphs with the prompt, “Explain this in terms that a layperson can understand.”
As for robot chefs doing the actual cooking, Ramirez predicts that “while they'll never completely replace humans, they're going to get better and better and more efficient.”
Planning out a multi-week cycle is central to an on-site chef's job, and Ramirez has been refining AI to help with that tedious, time-consuming process.
“That's something I always have as a chef, and it takes a lot of time,” he said, holding up an image of a room with Post-it notes covering the walls to create a five-week menu cycle. “What we were trying to do was create a nice menu with those items combined. We had to start over.”
So Ramirez's next question was, “How does AI enhance our potential? How does AI enhance our HI?”
Fine-tuning the prompts for effective AI responses is a human job, he found, and not always easy. “With AI, we designed hundreds of prompts over and over again. The system kept breaking down. We said, 'We want a five-week menu cycle for six stations in a college town in the Midwest.' We had all these guardrails in place, but they would just break. So we had to do some trial and error. We tried a little at a time and settled on creating prompts for one station, one week at a time.”
And humans still have to oversee the AI when planning menus like this: On a Mexican concept's menu, for example, the AI suggested “Taco Thursday” instead of “Taco Tuesday.” On the same concept, Ramirez suggested an entrée, but instead showed black bean soup and queso fundido, which are more like side dishes or starters.
The same goes for recipes: In one example, Ramirez asked the AI to change a recipe to a lower-sodium version, and the AI simply replied, “Less salt, please.”
Essentially, when using AI to create recipes, “you have to double-check because the AI will lie,” Ramirez says.
When it comes to nutritional or allergy information, AI needs to be monitored even more closely. “You can ask AI to provide that information, but the AI is often wrong,” Ramirez says. “You can also use AI to generate images for menu items, but these images are taken from thousands of pictures on the internet, sometimes a combination of recipes and photos from thousands of people, which is fine, but obviously fake.”
Ramirez shared his screen and searched for paella recipes and photos suggested by the audience. At first, the photos looked delicious, but there was something a little “off,” almost like the “uncanny valley” – a realistic yet unsettling phenomenon caused by subtle signals and cues received by the human subconscious.
One success story is fast-casual chain Velvet Taco, which used AI to create five unique tacos from existing food items, and it worked. “They're calling them 'AI tacos,' and it's a really clever marketing plan,” Ramirez says.
Overall, AI is something chefs should continue to leverage in different ways. “When AI first came on the scene, it was one of those jobs for people who have a creative process as a chef,” he says. “We thought creative jobs were safe. We were safe. AI was just going to be more technical jobs for doctors and lawyers. But AI is now creating amazing things that we never expected. It's scary.”
