December 3, 2023

We all know we are not all equal when it comes to kitchen skills or willingness to spend time in the kitchen. It is one of the main drivers driving the use of new AI developments to shape the future of the home and commercial kitchen. (That, and the apparent windfall that comes with it.)

The latest development comes from the University of Cambridge’s engineering department. After watching instructional videos, researchers taught an artificial intelligence-programmed robot “chef” to use the machine’s brain to make salad by itself.

Watching a person slowly make eight different salads, and by using mathematical formulas to help the machine translate visual cues, the robot was able to identify ingredients and prepare different salads based on its mental “recipe.” What’s more, the robot is able to build on its knowledge as it progresses and come up with a ninth recipe on its own. However, this original recipe appears to be just a reordering of the ingredients used in another salad.

“We wanted to see if we could train robot chefs to learn in the same incremental way as humans – by recognizing ingredients and how they go together in a dish,” said lead author Grzegorz Sochacki from the University of Cambridge’s Department of Engineering.

Recipes are limited to five ingredients: broccoli, carrots, apples, bananas and oranges. Three recipes are salad variations, three are fruit salads, and two are vegetables. It’s safe to say this robot chef isn’t quite ready for a wedding, or even a date night. Of the 16 instructional videos the robot watched, it identified the correct recipe 93 percent of the time, even though it only detected 83 percent of human chef demonstrations.

Sadly, AI isn't smart enough to know that you can't win friends with a salad
Sadly, AI isn’t smart enough to know that you can’t win friends with a salad

University of Cambridge

Of course, there are obvious limits to what a robot can do. In addition to being unable to perform the complex movements that human chefs typically require when preparing food, its neural network trained on the Microsoft Common Objects in Context (COCO) image recognition dataset is still very rudimentary.

However, the researchers were impressed by the way the robot chef was able to infer actions from the video; for example, if the demonstrator was holding a knife in one hand and a carrot in the other, the robot knew it had to cook the food on its own. Place the vegetables in the slicer while they are being prepared.

“Our robots weren’t interested in the kind of food videos that went viral on social media—they were too difficult to understand,” Sochacki said. “But as these robot chefs got better at recognizing ingredients in food videos, Well, sooner or later, they might be able to use sites like YouTube to learn all kinds of recipes.”

While automating food production is certainly not a new idea; it has brought us consumer electronics like the CookingPal and, more recently, culinary monsters like 3D printed “cakes”. While machine learning has the potential to be a game-changer for home cooks and professional kitchens in some ways, it’s safe to say that no front-line chef has put his head to the chopping block just yet.

This robotic 'salad' adds new meaning to the term minimalist cooking
This robotic ‘salad’ adds new meaning to the term minimalist cooking

University of Cambridge

“The recipes aren’t complicated – they’re essentially chopped fruit and vegetables, but it’s very effective in identifying, for example, two chopped apples and two chopped carrots versus three chopped apples and Three chopped carrots are the same recipe,” adding: “It’s amazing how much nuance the robot can detect.”

The study was published in the journal IEEE access.

although not chef Listen to the video and check out the Cambridge Robot Chef in this video.

Robot ‘chef’ learns to recreate recipes by watching food videos

source: University of Cambridge