Many of you have by now seen the Moley robotic kitchen, which was revealed two weeks ago at Hannover Messe. I was hired as a development chef for the project and ultimately became a sort of spokesperson for it. My role was, in fact, quite limited when compared to the work done by the robot programmers and technicians at Shadow Robot, the company that designed and built the extremely sophisticated robot hands that make human-like cooking possible. I was really just hired to cook and fine-tune the food itself – a few trial recipes to begin with, and then several run-throughs of the first dish we decided on: a crab bisque. Though I have been involved with the project since autumn of last year, I only spent a couple weeks in total working on it. It was the people at Shadow who really put in the hours, especially as we approached the launch date, when they seemingly never stopped working. Those guys all looked pretty damn tired by the time we got to Hannover.
But despite my rather limited role and knowledge, I have been approached for a lot of interviews about the robot over the past couple weeks. I suppose this is because most journalists are more interested in the implications this has for the future of both professional and domestic cookery than they are about the technical aspects of the robot, which are quite… technical. I have noticed some of the same questions cropping up again and again, so I figured I’d address them myself here. This is also to clarify and correct some things that have been misreported entirely (for example, Engadget said that I designed the robot) or otherwise needlessly editorialised with uninformed speculation or testimonies from people who are not involved with (and have never even seen) the robot.
I should also note that these are not responses fed to me by a PR person. These are my own perspectives on the robotic kitchen that I’ve arrived at having seen what it can do, and having thought about these issues quite a bit.
How does it work?
The robot cooks using only pre-programmed movements; it has no sensors of any kind. Everything the robot does must first be done by a human (in this case, me) to ‘teach’ it how to move in certain ways. To do this, the robotics team set up two identical kitchen layouts in their workshop, one for the robot, and one for myself. With the ingredients and utensils laid out in precise locations on the worktop, I strapped on some motion-capture gloves and wristbands and cooked through the recipe as 3D cameras recorded my actions and automatically transposed them onto the robotic hands across from me. For the most part, I cooked naturally, as I usually would, but I had to make slight changes to my movements to accommodate the robot. For example, its grip is different from that of a human, so I had to hold a spatula differently from how I would normally, and I had to remain within a set area within the work station, outside of which the robot could not reach.
After recording the recipe five times, my work was essentially done until several months later when I had to come back and adjust the bisque’s flavor. But in those several months, the robotics team put together the smoothest, cleanest, most accurate movements that I made and tweaked them, often down to the millimeter, to ensure that the robot’s actions were always successful, since it had to operate without sensing or human guidance. And now we have a robot that will make an excellent crab bisque, using only rote actions, every time.
So it only makes a crab bisque?
Yes, for now. Bear in mind that what has been launched in Hannover is a proof-of-concept prototype, not a finished model for consumers. However, the process of making a crab bisque contains important actions, such as stirring, dosing, blending, and ladling, that could be remixed to create dozens, maybe hundreds more dishes. For example, if you were to stop the crab bisque program midway through, you would have a simple pasta sauce of butter, shallots, garlic, cherry tomatoes, and vermouth. If you were to use chickpeas instead of crabmeat and a blend of spices instead of salt and pepper, you could produce a chana masala.
There is obviously an enormous amount of movements that have to be programmed into the robot’s repertoire to enable it to cook with anywhere even close to the same range that a human has. But even with just the same movements that go into making a bisque, there is the potential to make a very wide selection of other dishes.
How is it different from previous examples of kitchen automation?
For the most part, kitchen automation in the past has taken the form of machines that only do one thing, or just a few of things. The most rudimentary things we might call ‘kitchen robots’ are perhaps rice cookers or toasters. Modern rice cookers often use computer-assisted moisture and temperature sensors to automatically adjust heat and timing to cook rice in a way that corrects variables like starting temperature and how quickly the rice absorbs water. Similarly, state-of-the-art toasters use light sensors to gauge the color of the bread inside, producing toast that is evenly and consistently toasted to the operator’s liking. Both of these are pretty cool. But ultimately, they only do one thing: make rice or make toast. Far more sophisticated examples would be the Japanese okonomiyaki and ramen robots, but although their movements are more impressive and the end products are far more complex, they are still only really capable of doing one thing.
The crux of what makes the Moley robot innovative and groundbreaking is the use of robotic hands. The robotic kitchen is the brainchild of Mark Oleynik, who was very astute in his assessment that human hands are the most important cooking tool we have; we use them for everything. So Mark’s stroke of genius was to use robot hands, which are potentially limitless in their capabilities, instead of a range of gadgets that are only capable of a few different tasks. Of course, the existing robotic hands are not quite as subtle or responsive as actual human hands, so there are still limitations. But personally, I believe they can be overcome. One chef who was (inexplicably) approached for an interview about the robot kitchen defied it to make sushi, claiming it is too delicate a process. But actually, sushi robots already exist, and robotic hands will only improve the quality of sushi produced through automation.
So if it can do anything and everything, does this mean it will replace professional chefs and home cooks?
Mostly no. (But possibly, in certain situations, yes.)
On the domestic front, this will not keep anyone from cooking when they want to cook. It isn’t intended, nor do I think it will be used, as something that will eliminate home cooking. What it is meant to do is give people a new option for when they cannot or don’t want to cook; if it replaces anything, it will be ready meals and takeaways. I love to cook, but I don’t always love to cook; sometimes, I’m too tired or too busy. The robotic kitchen would help in these instances, and would not detract from others. Whereas I may now resort to poor-quality food that has been pre-made and packaged, or delivered from a questionable local restaurant, in the future I could have the robot make a good meal from fresh ingredients. The kitchen is designed for human use first, and robot use second; it will allow you to cook when you can, and be cooked for when you can’t. I should also point out that the robot can collaborate in the kitchen; it can cook part of a recipe, allowing a human to do the rest, and it can also teach human cooks new techniques for them to use in the future.
As for professional settings, there are a number of reasons why the robot probably won’t replace working chefs in any significant way. Professional cooks rely on very human attributes to succeed in their job. They must constantly look, smell, feel, listen to, and of course taste what they are cooking in order to assess the food and adjust their actions accordingly as they go. This is down to the irregular nature of ingredients and the speed at which professional kitchens move. There is no time for a chef who didn’t cook an onion properly because it failed to understand that it didn’t chop it right in the first place, and then wasn’t able to intuitively account for differences in that onion’s size, sugar content, moisture content, etc. that may cause it cook (or burn) at a different rate. Professional kitchens are places where quick reactions to new, surprising situations are crucial; this sort of intuition will be beyond the robot’s abilities for a very, very long time. There is, of course, capacity for the robot to take on some of the repetitive drudge work in the kitchen, like basic prep and cleaning. But even these tasks require an ability to see and understand what is happening; dicing a banana and washing dishes will are basic tasks, for sure, but it won’t be easy for the robot to comprehend when it has just chopped a bruised banana or missed a spot of egg yolk on a plate. I believe the robot can and will be used for certain jobs in professional kitchens, but they will arrive so gradually and in such a minor way that any job replacement will hardly be noticed.
And even if the robot can be made just as intuitive and dextrous as a human cook, there are two reasons why it will never come to replace humans entirely. Firstly, it cannot create new dishes. To be able to combine or recombine ingredients and techniques, to improvise with flavors, and to gauge the seasoning and balance of a finished dish is beyond the robot. This means, at the very least, we will always need humans to create dishes and ‘teach’ them to the robot. Secondly, humans value things that are handmade because they are handmade. There’s an awful lot of things that could be produced entirely through automation– everything from beer to furniture –but they aren’t. Often, this is because the end product won’t be as good, but possibly more often it’s because we appreciate the idea that what we consume has been produced with craftsmanship, knowledge, and passion, by a human being.
I may be wrong about all of this. Perhaps someday the robotic kitchen will be so sophisticated that its output and creative capabilities will be indistinguishable from that of a human. But having witnessed firsthand what the robot can and can’t do (albeit at a very early stage), this seems extremely unlikely.
Is it safe?
Yes. A few articles have insinuated that if a pet or even a human were to get in the way of the robot, it could be harmed or killed. This is simply untrue. Even in its current, prototypical form, the robot kitchen is outfitted with a plexiglass shield that closes in front of the unit to keep out curious cats or wandering hands so they don’t get in harm’s way. When the unit is launched for consumers, the robot will not operate unless this shield is closed, nor will it operate if it detects any foreign object in the kitchen. It will be no more dangerous than a food processor. Maybe less so.
If you have any other questions I haven’t covered, please do leave a comment!