For a chef, receiving a Michelin star is the equivalent of winning an Oscar or an Academy Award. The Michelin Guide’s awarded Michelin Stars represents a lifetime of hard work and has become the most coveted and respected culinary award a restaurant or chef can receive. In order to be considered for a Michelin star, a restaurant must show a mastery of ingredients and cooking techniques and showcase the personality of the chef in the cuisine. With only a handful of chefs deemed worthy of this distinction, they have all shown a tremendous amount of creativity and technique mastery. One of the reasons why earning a Michelin star is so difficult is that these critics want to have a completely new experience.
As a chef, it is really difficult to create and innovate simultaneously making sure that it is a flavor that is not too familiar. A chef’s signature cooking style and menu curation stems from their cultural background, where they were taught and trained, and even the flavors that they grew up around. For example, as a chef from the Philippines, I am familiar with using sweet and savory combinations in dishes. I also grew up with a lot of Laotian and Vietnamese influence so working with fish sauce is normal for me. With these flavor backgrounds, I have my own personal styles to cook with but it is not innovative at all. In this week’s edition of weekly blogs, we will be talking about creative computing and how chefs feed AI algorithms recipes in order to find patterns and create new dishes.
IBM and Chef Watson – Early Applications
One place that I did not anticipate was the integration of AI in recipe creation. In 2012, teachers from IBM collaborated with the Institute of Culinary Education (ICE) and created Chef Watson to apply computational creativity to the culinary world. You may be wondering though – “how does a machine with no sense of taste or smell, be trusted to create new recipes?”. The creators of Chef Watson fed the program with around 10,000 recipes from Bon Appetit’s archives and utilized natural language processing to analyze the recipes and learn the logic behind ingredient usage and how they were combined. What sets Chef Watson apart from human chefs was their ability to understand complementary flavors. A chef on average can only think of pairing a few ingredients together and that is only because of years of experience working with those flavors. Chef Watson has the capacity to look at pairings of six, seven, or even nine ingredients at a time without a problem. This is the reason why Chef Watson was able to advise cooks and discover totally unique recipes using this flavor compound algorithm. The professional version of Chef Watson was used in the ICE test kitchen which consisted of extra thirty-thousand recipes and added information about the molecular makeup of smells and flavor compounds. In addition to the extra information in the professional version of the program, IBM also fed the program information regarding hedonic psychophysics that investigates the smells and tastes that people tend to enjoy.
How Chef Watson Works
Chef Watson is a computer program designed to help cooks discover and create original recipes with the help of its flavor compound algorithms. A cook starts by choosing an ingredient they want to cook with from the list of options. Then, the cook selects what they are actually trying to make, whether it be a pasta dish, paella, burger, etc. Finally, the cook selects a theme in which they want to cook. This can be anything from “Chef’s Day Off” to “Tuscany Nights”, this just represents the ambiance in which you would like to consume what you plan to make. After all the prompts are answered, Chef Watson generates a list of recipes that would be good to try that fit the parameters that the chef specified in the program.
(Note: ICE and IBM Published a cookbook with 65 recipes made by Chef Watson.)
Sony AI and the “Gastronomy Flagship Project”
In 2020, Sony AI began working on its own recipe creation app. In addition to recipes, Sony AI will utilize data sources such as flavor and molecular structure. Similar to Chef Watson, Sony AI wants the app to supplement the chefs in their creative processes including ingredient pairing, recipe design, and menu curation. This is just one arm in Sony’s Gastronomy Flagship Project, which aims to study how AI and robotics can help in every aspect of culinary work, not just within theoretical menu creation and curation. Sony also wants to develop robotics that can assist chefs with the preparation part of the culinary adventure. The goal of robotics is to learn the various preparation techniques it takes to complete certain tasks. This may be as simple as understanding the proper terminology of cutting and preparing ingredients to cooking techniques. Sony AI aims to create a solution that can assist chefs through the entire process, from preparation to the plating.
Everything in the culinary world is about experimentation and feedback. The third arm of the Gastronomy Flagship project is focused on building the feedback aspect of the culinary journey. Sony AI has announced an interview series where chefs of various backgrounds are interviewed to speak on their creative process, sources of inspiration, and their use of technology. Sony AI is still actively looking for more partners in order to continue its research and refine its models. Chefs such as Chef Hajime Yoneda, owner of 3-Michelin Star restaurant HAJIME in Osaka, Japan one of the collaborators working on this project.
Conclusion
Recipe creation is the backbone of creativity in the culinary world. Chefs derive most of their identity through their social, ethnic, and educational backgrounds, which all translate to the way that they create recipes. Programs like Chef Watson and the Gastronomy Flagship Project use natural language models along with data from recipes, smells, tastes, and molecular structures in order to find patterns in recipes. Feeding these programs with recipes from award-winning chefs will allow the programs to learn the patterns that make each chef unique while removing the biases and preferences that come naturally to each individual chef. This program, combined its creativity with a chef’s technical expertise opened the door to the discovery of new flavor pairings, never before enjoyed by diners to this day. Who knows? Maybe a few years from now, when you get the chance to sit at a fine dining establishment, you would not even know that the dish that you are enjoying was based on a suggested recipe by AI.
Hey Fernando! I think it’s awesome that you connected your experiences to the topic! The fact that they can put in some ingredients they would like into the system will put in what ingredients will go well together with regional cuisine. Some may say that you aren’t a real chef if you use this tech, but isn’t that the whole idea of pushing the boundaries of creativity and the evolution of our society?
Hi Fernando! This is such a unique concept and a really good blog post idea! I think that using AI in something like cooking can help chefs eliminate some of the concern related to wasting time and resources. It’s interesting to see how the AI will choose ingredients to pair and recipes based on what the chef is looking for, but like you mentioned a lot of cooking is related to background of the chef. I wonder how accurately the AI will be able to mimic traditional recipes. It’ll be really interesting to see. Also ever since the pandemic started, I’ve seen more food delivery services with recipes and ingredients that will be sent to you based on food preferences that the person chooses, so something like Chef Watson can also help the people who use the subscription services and they can also try many more varieties of recipes. Great job!
Hi Fernando, I love this topic for your post. I thought about how artificial intelligence could be used in fast food restaurants with placing orders and flipping burgers, but I did not think about how it could be use to form recipes based on complementary flavors. I am curious to see how much Chef Watson will develop as it is being programmed with recipes from some of the best chefs in the world. Also, it will be interesting to see how many restaurants, including fine-dining restaurants, will incorporate recipe suggestions by AI into their cooking. Overall, I really enjoyed reading this post!
Hey Fernando! You combined two of my favorite things in one blog post, food and tech! It’s really cool to learn about how Chef Watson and Sony AI are leveraging data sources such as flavor and molecular structure to create unique recipes and aid in chefs’ creative processes. It’s interesting to see how AI is being used to help chefs create new and innovative dishes while maintaining their unique cooking style and personality. And helping win Michelin stars! Great post.
I love this. I also agree with what Connor said about this not being a cop-out, but pushing the boundaries. Because as you mentioned in the article chefs can only think about so many flavors at one time, so why not maximize that and free up your time to then actually craft the dishes? This sounds like a perfect application of AI if functional.
This technology is so cool! I agree with the comments that this is a practical application of AI to find new patterns and combinations of flavors that a human couldn’t detect with their limited exposure to the mixing of ingredients. I think this technology would also be interesting to implement for personal at home use through the creation of an app that shows unlikely, but tasty, pairings of everyday household grocery items. Great post!
Cool post and tying it back to your roots of being a chef! It’s interesting that IBM and ICE decided to name the chef after the founder of IBM, Thomas Watson. I went to Youtube looking to find someone taste testing any of the 65 recipes by Chef Watson. To my surprise, nobody has ever done it or at least there weren’t any immediate results. I did find a video from 2 weeks ago https://youtu.be/ZNmX4jtpo2A and it’s a human chef competing in a challenge against a AI chatbot recipe. Great watch
Cool post, i always find it fascinating when people are good at cooking, I am horrible at it and still struggle making a grilled cheese. It is interesting to see how technology can help chefs in their creative processes, from ingredient pairing to recipe design, and even to the plating of dishes. Chef Watson and Sony AI’s recipe creation app are excellent examples of how AI can help chefs to experiment and discover unique flavors, and I look forward to seeing more of these technologies in action. Maybe in the future I’ll have some AI assistance in the kitchen because the online recipes don’t seem to be working for me.
Fernando, awesome post! The way that AI can bring innovative dishes and and flavors beyond the bias of the chef is incredible. I guess I’m a little more old fashioned but personally, I think the human aspect of creation is what makes food real and genuine. However, if one day I am given a meal that’s recipe was AI originated, it would change my mind. One thing that I thought of while reading this is how AI given genetic data about a human as well as molecular food data, could generate meal plans and recipes tailored to health and longevity.
Hey Jorome! It is so interesting to see AI gather a plethora of information about recipes to curate new ones! I can imagine how helpful this is in developing and experimenting with new cuisines and ingredients of food. If I were to sit at a restaurant and find out that the food I ordered was suggested by AI, I would be blown away. The more data we feed AI, the more it can feed us!!
Hey Fernando! I enjoyed reading your post. I find it very interesting to see that AI can help generate an incredible recipe that could be used by a chef, or by anyone in general.
Hey Fernando, I found this post super interesting to read. This post reminds me of another blog post from earlier this semester about AI in the brewing industry. It would be awesome if in the future people can have their taste buds scanned and sent personally designed recipes.
Hey Fernando! It’s really cool seeing all of the possibilities of AI. One of the biggest things that I thought about, similar to Jhanvi, is how AI would generate recipes for traditional meals. As a West African, a lot of the foods we make consist of a very interesting combination of flavors. I wonder if AI would be able to mimic or even help improve those dishes.
Great post fernando! this is super cool technology and I had no idea its been around since 2012. I feel as if this would be a great idea for a new or young chef to look into in order to get familiar with well known flavor pairings but even could be used to find entirely new ones. Its also cool that Sony’s version looks at molecular structure in order to find complementary recipes. Sometimes I forget that cooking is a science and this technology proves that.
Hey Fernando! This post was very insightful. I would love to have the computer program named Chef Watson accessible to myself, so I can test out my skills in the kitchen. I do believe a few implications arise from this program however. This does take the creative process away for chefs and eventually with enough data AI could have every single recipe possible. I am unsure if this is a good thing or a bad thing. Nonetheless, I enjoyed reading and thank you for sharing.