Trainable and subjective devices for the future kitchen.
Looking back at the history of cooking tools and kitchen appliances, machines in the kitchen have defined new cooking processes and have automated parts of the act of cooking. From the invention of the blender in the 1920s to the recent smart ovens and robotic-cappuccino-makers, tools are designed to make cooking more convenient by outsourcing labour-intensive tasks to machines. In recent years, with lowered cost of computational power and easier access to robotic technologies, we see more visions of fully-automated kitchens in restaurant experiences or even home solutions.
As part of our first lab, Kitchen Intelligence here at YEAST, we looked at a different route for the future, exploring machines that focus on collaboration instead of automation, and where training and learning are new ways of interacting with a device.
AUTOMATION, ROBOTS AND THE ACT OF COOKING.
Recently, there have been examples of ‘first’ robotic restaurants, bars and cafeterias. Creator, Spyce, and MakrShaker showcase robotic or robot-like machines that can make burgers, stir-fry noodles and prepare coffee or a cocktails for you. A less polished version of these machines can be found often in China, where a robotic noodle shaver has been released in the wild many years ago, in a variation of uncanny forms.
Some of these examples use the wow-effect of people seeing robotic arms, which are the same robotics used in heavy industries, like car manufacturing, but used for way less complex tasks like, used to flipping burgers or mixing martini cocktails. Some instead use industrial-like production line concoctions, that once used to be hidden in factories, but now are redesigned and shown very openly to push for what they define as a new culinary excellence enabled by technology. These examples definitely serve the purpose of opening the discussion about automation in food experiences and to gather attention and investment. However it also raises questions around the future role of people in the kitchen, and in workplaces in general. As the first flipping burger robot was recently fired, it also raises issues about future of work for robots themselves.
The role of people in this new robotic kitchen future is, in the best case, left to the final bit of the act of cooking. In cases like Creator or Spyce, the machines/robots do the most work and human workers are in charge of decorating the plate and keeping the social interaction with customers. This could be seen as a positive move up in the value chain of the kitchen for workers, but it could also mean offloading big parts of the act of cooking to machines.
When talking to chefs in our previous research what least resonated and often scared them was to imagine a future where the manual act of making the food would be automated. The manual act of cooking, cutting, garnishing is often where chefs feel their ideas, knowledge and emotions passing through in their dishes. The difficult-to-measure additions, pinches, adjustments are the difference between an industrialized dish and a crafted one. Of course if this is framed as a ‘problem’ or as a ‘chore’ it makes sense that a machine could do this job instead of a human. But borrowing from Cedric Price: if robots in the kitchen are the answer, what is the question?
A machine preparing food is likely to be trained by someone to make a ‘perfect’ burger or a ‘perfect’ cocktail and it will also be able to potentially iterate faster through variations to ‘learn’ what is most ‘successful’. But as we have realized in more objective and impactful situations, such as algorithmic security or self-driving cars, algorithms in automated systems are far from ‘perfect’ and are prone to mistakes and biases of what is defined as ‘successful’ and what is ‘learned’ from wrong or limited datasets.
While we want to evolve cooking to reach new heights and imagine the future of cooking, is flipping burgers really the best use of a highly precise robotic arm?
LEARNING AND TRAINING THE IMMEASURABLES OF COOKING.
Looking at a different direction for the future, we have explored how to design devices that would reinforce the idea of collaboration between people and machines in the future kitchen. Rather than focusing on replacing people with big robotic arms, we looked at smaller and more approachable ‘robots’. Instead of designing something to achieve a ‘perfect’ answer or recipe, we focused on retaining or augmenting the knowledge of the person in the act of cooking.
As more machine learning algorithms are and will be more and more embedded in everyday objects, it will change the way we interact with them. A device will need to be trained by you or adapted to your routines, habits and way to interaction. What is also interesting about this is that a device could be trained by you and for you only, becoming a truly personal and subjective thing.
At a workshop at Copenhagen Institute of Interaction Design, together with Massimo Banzi, we explored how training and learning could be a metaphor to change not only the way we interact with products, but potentially also their value. With a group of students (Micol Galeotti, Alex Penman and Sareena Avadhany) we explored how a chef’s knowledge could be embedded and learned by his tools and how that knowledge could be passed on to others. By recording the movement of cutting various foods, a knife that learns from a professional chef could embed and teach the knowhow and subtleties of cutting different things, based on the chef’s experience.
As we explored more collaborative and trainable devices for the future of cooking, we started looking at more of those immeasurable and subjective decisions that in the act of cooking. We looked at actions that can be trained and learned by a device from an expert, and also how to keep the memory of what you consider ‘perfect’ and ‘good’.
As the more quantitative parts of cooking (grams, seconds,…) are already being recorded and learned by devices to cook more consistent meals and recipes, there are however certain measures and decision that are difficult to define and subjective. Think of the ‘pinch’ you add to your seasoning, or what you consider the ‘precise’ way to cut a vegetable for a stew, or the ‘bit more’ of time that you cook pasta for to get to the right ‘al dente’.
We worked together with Chiu Chih to prototype three different devices that using machine learning tools (Arduino 101s, Raspberry pi and Wekinator) would learn from chefs and non-chefs what is considered a ‘pinch’, what would be considered a ‘precise’ cut and what is their own measure of ‘a bit more’ time to them.
Trainable Measures is a series of kitchen helpers that are trained to help you define subjective measures in the kitchen. Everyone’s ‘pinch’ or ‘a bit more time’ to cook are unique and subjective measures to you, as ‘precision’ in how we cut something is hard to measure. Rather than create machines to fix and define ‘perfect’ measures to adhere to, we use training and learning as a way to let people create their own measures. Trainable measures can be used by you to learn and create your own measures overtime, or can be trained by expert chefs to pass some of the intangible and often hard to quantify details that might make a difference in the kitchen.
THE ‘PINCH’ MACHINE.
The pinch machine is a small robotic arm that can pinch salt and spices and release it on a plate or a pot nearby. It can be trained to replicate the amount in weight of salt or any other spice that you might need a ‘pinch’ of. It is a combination of a weight scale and a small robot ‘arm’ pincher. By taking several times a pinch of salt, the pinch machine will try to pick up a similar amount of salt. Everyone pinch is different and the machine itself might have its own ideas too.
THE ‘PRECISE’ SORTER.
The precise sorter is a small sorting device with a camera looking down at the cutting board. It can be trained to recognize whether something is cut correctly or not. You can train the ‘precise’ sorter each time by showing the example of what is precisely cut and what is not, and as you push things to it will sort them for you. As precision in cutting is often at the base of a great dish, the precise sorter can help you create consistency and improve your dishes.
THE ‘BIT MORE’ CLOCK.
The ‘a bit more’ clock is a timer for that bit more you need for cooking something ‘perfectly’. The clock defaults with 60 seconds as the measure of “a bit more”but based on how you actually use it, the measure of “a bit more” changes over time. Over time, the clock will show you your own subjective minute or ‘a bit more’.
YEAST is a future venture laboratory where we imagine, build, and run companies that improve living through food and technology.
Kitchen Intelligence is a lab where we research and experiment using artificial intelligent technologies in the context of commercial kitchens. We seek to build ventures where technology augments rather than automates the abilities of chefs.
This is part of the series for Kitchen Intelligence, an YEAST lab.
- Kitchen Intelligence: Introduction to a YEAST lab.
- Kitchen Conversations: A look into what meaningful roles technology could play in commercial kitchens.
- Ghost Chef: Augmenting and supporting creativity for the future kitchen.
- Trainable Measures: Trainable and subjective devices for the future kitchen.