As we introduced previously here, many technology companies today envision a future in which AI technologies and machines fully automate the cooking experience. These visions pitch for greater production efficiency and consumer convenience in ways that humans don’t have too much involvement in the actual cooking, if at all.
At YEAST, we challenge this vision. Technology for automation is solving for the wrong easy. Certainly a great deal of automation will continue to enter our everyday lives, as it always has, but that will only create greater opportunities for creation by humans. The role of technology vs human in the creation process is the important question.
COMPUTATION, CREATIVITY AND KITCHEN KNOWLEDGE.
Machines have played a role in simplifying the ‘manual’ labor of chefs for years: in recent history, for example, sous-vide technology has helped chefs reach unprecedented levels of precision. But is manual labor what chefs desire from future tools and application of technologies in the kitchen? Certainty not for many common home cooks, but even at the high end of fine dining, precision is not the most valuable application of technology.
Our research insights (more here) pointed us in two directions that technology could play a more meaningful role in the everyday life of chefs: augmenting creativity and capturing potentially lost memory and knowledge. Technology can have a game-changing role in supporting chefs to deal with highly dynamic and complex sets of information.
We talked to chefs about creativity in cooking and the types and ways of using different tools in their creative process. Most of these tools are analog and archaic; of the digital tools used, most are basic in function and general in design — none were designed specifically for chefs or cooking. There’s really no comparison to the creative tools available to designers and engineers, for example. In these fields, digital tools reinforce the creative processes by handling large amounts of information and by generating options to inspire new ideas. The value of computational power for creativity is enjoyed by many, but not chefs.
Even if the work of chefs end up being tangible and actually edible things, digital and generative tools could transform the way chefs imagine, create and prepare food in the future. How could AI technologies enhance creative processes during the creation process? How could computational power help chefs parse through large ingredient databases to curate interesting sources of inspiration? How could machine learning help preserve and then extend the knowledge of a kitchen?
BEYOND RECIPE BOOKS AND CO-CREATING WITH A MACHINE.
An essential part of a chef’s training comes in the form of apprenticeship. Learning from other chefs, through observation and then by trial-and-error, is how practical knowledge and techniques are learned. Creativity is then inspired from this foundation, either to evolve or disrupt what was learned from before. How could we recreate a digital version of this apprenticeship experience?
When chefs learn from other chefs, it’s not just about recipes — chefs don’t read recipes. If a chef browses a recipe book, they are seeking to understand the overall approach, preferences of ingredients, peculiar techniques, and then with that collective perspective, they extrapolate to their own cooking. Highly trained chefs are able to understand the principles and approaches from looking at the final recipe in just a single picture.
So we wanted to build a way to better emulate the learning experience of chefs by enabling collaboration with other chefs. We also needed to recognize that recipes are better used to understand principles and processes. To that end, we imagined a co-creation tool that would help craft suggestions for recipes not by generating an outcome, but rather as a conversation between two chefs.
Our initial experiment was done by writing recipes simultaneously with chefs from around the world on Google Docs. Chefs were asked to simulate the intelligence inside of an imaginary plug-in tool on Google Docs. From this initial experiment, we found it quite natural when chefs would interact with us by “auto-completing” part of the recipe as we were typing mid-sentence, or by generating a list of options for us to choose from before we proceeded to the next step in the recipe.
This experiment inspired us to build a prototype of a tool which we call Ghost Chef. Ghost Chef was designed and developed in collaboration with Lorenzo Romagnoli and in collaboration with Bits x Bites.
GHOST CHEF.
Imagine you can cook a gazpacho with Ferran Adria making suggestions, or compare how a Nordic and a Japanese chef would pair salmon differently. We imagined Ghost Chef to be a conversational tool that inspires chefs to create new dishes, rethinking what the future of recipe books could become.
GhostChef is a chef learner tool, a minimalistic recipe writer that allows professional and amateur chefs to type and save their recipes, while the platform learns the chef’s ingredient combinations, preparation steps and cooking methods to help chefs build a GhostChef culinary identity that can be shared with others.
GhostChef is a chef creative tool, allowing users to get inspired by cooking with other chefs, experimenting and exploring recipes by collaborating with digital identities created by a global network of chefs from kitchens as diverse as fine-dining restaurants and street food stalls.
NARROW INTELLIGENCE.
GhostChef is not a general intelligence for all food recipes; it is a context-specific intelligent tool built atop a narrowly defined dataset generated by chefs themselves. Users select different digital profiles of chefs to write recipes with, each of which generates a distinct and interesting set of suggestions. The possibilities are not comprehensively exhaustive, but they are non-generic and specific to a particular chef. In other words, you get suggestions as you would from another chef.
ADJUSTABLE INTELLIGENCE.
Users initiate the interaction by typing an ingredient, presumably what is available or desirable to the user. Generated suggestions are adapted to actual user inputs, and suggestions are given with at least two options, ensuring control for relevance and desirability. Meta-data adjustments also filter and prioritize the suggestions based on a limited set of variables including price, complexity, speed and locality.
YEAST. is a future of food laboratory. We explore the relationship between food, emerging technologies, and urban living.
This is part of the series for Food x Machines.
- 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.
