Can a Robot Learn to Cook?

Everyone’s coming over to watch the big game. You’ve got beer, a giant high-definition television, and a well-deserved reputation for serving wings hotter than Dante’s eighth circle of hell. Unfortunately, you are pressed for time. Wouldn’t it be great if a machine like Rosey from The Jetsons could quickly prepare them? Maybe you could even pass off the dish as your own!

Then again, maybe not. Would Rosey’s version taste like yours, or would her rendition expose your duplicity? Could she cut the chicken into the right size parts and ensure your friends don’t choke on bone chips? Would Rosey know when the chicken pieces hit the ideal state of crispiness without being raw inside? Most importantly, could she discern when the spice Rubicon was crossed? These questions all revolve around one issue: Can Rosey can acquire tacit knowledge?

Contemporary discussions of tacit knowledge owe a debt to Michael Polanyi, a scientist and philosopher who famously observed: “We know more than we can tell.” Use chopsticks? Go mushrooming or fly-fishing? Know how to pick the right bottle of wine, or how to be a charming dinner guest who tells the right jokes and not mortify your host? These skills use tacit knowledge. It’s what enables us to appreciate what something means, given context.

While we all possess lots of tacit knowledge and couldn’t conduct our daily affairs without it, experts use it as their secret weapon. In Blink: The Power of Thinking Without Thinking, Malcolm Gladwell described how experts rely on intuition. Without deliberating over formal decision-making procedures, they can look at a problem and immediately have the “aha” moment that reveals the solution.

Although this ability is immensely powerful, it isn’t magical or innate. After prolonged and committed experience, experts develop superior pattern-recognition ability that lets them see, at glance, how new situations resemble remembered ones.

Experts aren’t unique in this regard. All humans require extensive hands-on experience or social interaction to develop tacit knowledge. As every cook knows, mechanically following a recipe will only take you so far. Nevertheless, Gary McMurray, chief of the Food Processing Division of Georgia Tech’s Research Institute, believes robots will acquire the tacit knowledge needed to debone and butcher a chicken.

While many of us had a parent — or other mentor — guide us toward the basic tacit knowledge required to cook, the secret sauce for his project is a 3D imaging system that quickly measures and calculates the dimension of each individual chicken before the robot carves away. How can his “intelligent cutting” project lose? It has the support of custom algorithms underwritten by complex mathematical equations.

McMurray’s team realizes they have to solve many adaptive challenges. After all, every chicken has a different muscle structure, different joint segments, and carries a different weight. Despite this complexity, an experienced human chicken processor can butcher 1,000 chickens an hour, immediately judging how and where to cut without wasting meat. And here’s where the difference between person and machine becomes salient. Shannon Heath, spokesperson for the world’s leading chicken processing equipment, Meyn Food Processing Technology BV, says that technology “can’t quite match how closely a human can cut to the bone.”

So far, McMurray’s team has experienced success with the key technological components. Significant as this victory is, it remains a far cry from developing a fully integrated system, which, itself, wouldn’t be the final step. Vangelis’s Chariots of Fire musical score shouldn’t play until the robot is rolled out of lab and into the real world, where new skills might be needed. Still, McMurray told us he’s optimistic about the future and even is considering making commercial devices.

For the sake of argument, let’s say McMurray is right and it is only a matter of time before technology cuts chickens with the same speed, dexterity and accuracy as humans. Indeed, glimpses of the future are already here. Think someone hand shredded your mozzarella? Nope, a machine did. China’s got a series of noodle-bots that can hand-slice noodles into pots of boiling water, and the Japanese have used mechanic sushi makers for years. But will technology ever replicate the deft touch of China’s best noodle pullers? Or, Jiro Ono, Japan’s 85-year-old, three-starred Michelin sushi master?

For most chefs, automation per se doesn’t bother them. According to John Regefalk, chef at Rome’s Metamorfosi: “Surely most of the mechanic [sic] tasks in the kitchen could be executed by a robot, some of them already are today.” That said, he maintains: “Other things that will be very hard to implement in robots is the ability to make decisions based on personal judgment. Would a robot be capable of determining if the fish delivery is fresh enough or it needs to be sent back?”

Carol Choi, former pastry chef at Noma in Copenhagen agrees: “I do think that one day when technology gets advanced enough, robots could make the same dessert, but will they know exactly which pear is exactly ripe? Would they be able to see a plate and adjust for each individual element on the dish?”

Where does this leave our robotic chicken hacker, especially since the reach of robots surely will continue to grow? While Regefalk and Choi emphasize the difficulty of automating aesthetic judgment, Chris Selk, Executive Chef at Borgo Trattoria in Canada, gets downright existential, declaring: “It is human touch that gives ‘life’ to these ingredients, and an element of human fallibility that lends value.”

While Rosey might be able to make some kind of chicken snack, it is hard to imagine she’ll ever know how sublime the wings and a game experience can be without being able to eat. Sociologist of science Harry Collins hits the nail on the head in Tacit and Explicit Knowledge: “The growth of automation has provided new problems and more demanding question about what knowledge might be even though it remains the case that, in the last resort, humans are the only knowers.”

While, as Collins shows, embodiment might not be the key for acquiring all tacit knowledge, it has purchase here. An ignorant robot palate is our culinary loss, and mom’s chicken soup wouldn’t mean so much if she snuck us a Rosey-made bowl.

The Atlantic