in

Need a mouth-watering Thanksgiving recipe? Ask AI

What do you bring to Thanksgiving when your aunt is gluten-free, grandma needs to stay away from sodium, and your brother has a dairy allergy? Finding a dish that can meet everyone’s dietary restrictions while also tasting delicious is a challenge. 

Innovation

Sony AI believes that in the near future, AI can help. 

Sony AI is pursuing a digital canvas that allows rich conversations between AI systems and chefs, helping them go even further with their creations and develop new and unique recipes. Their research in the area of gastronomy aims to one day change the way chefs create food combinations, pairings, and platings and assist chefs in their process of developing new, original recipes that are also healthy and support sustainability for the environment. That includes ingredient pairing and recipe creation.

It’s an interesting use case for AI, given gastronomy’s position, straddling both science and art. To learn more about the effort, I reached out to Dr. Michael Spranger, COO of Sony AI, to discuss AI technology’s methodology and future applications in the world of gastronomy.

GN: How did Sony AI first become interested in gastronomy as a use case for the technology?

Dr. Michael Spranger: Simply put, gastronomy is a fascinating area of unchartered territory for Sony. Just like Music, Films and Games, we see Gastronomy as a global entertainment business, with technological advancements constantly contributing to its progress. Kitchens over the years have evolved with new technology, and chefs bring some of the purest creativity to that technology.

Our aim is to develop technology, and more specifically AI and Robotics, for chefs that empowers them to be even more creative in developing delicious dishes while also helping them to drive issues such as health and sustainability. 

In the area of recipe creation, we think that AI can assist chefs to explore vast amounts of data associated with food, including existing recipes, chemical and molecular composition and other factors like nutritional or environmental impact data to create a new dish. With robotics in the kitchen, we are hoping to assist chefs in their cooking process, from food preparation and cooking to serving and plating. Neither of these research areas are easy to solve, and that is exactly why we felt they were appropriate for us to set as a grand challenge in our Gastronomy Flagship Project.   

GN: There’s a perception that AI and automation are replacements for humans. How do you envision chefs working with the technology, and what has the reaction been so far?

Dr. Michael Spranger: Sony AI’s initial focus is on high-end restaurants and their chefs. Our role is to enhance what chefs do already, driving creativity by thinking about, for instance, how to use algorithms to put more data into their hands during recipe creation and conceptualization; how to use robotic systems to elevate the quality and quantity of what is possible in the kitchen during a dinner rush; how to deliver human and robot collaboration at plating, to enable previously impossible designs of dishes. Creativity is not an easy route for doing things — just look at El Bulli, the world’s best restaurant that had to close for months of the year to develop its groundbreaking food — and we think we can play a role supporting it. 

A good example of this is in recipe creation itself. A challenge we have is to understand how far we should be taking recommendations in recipe creation. It doesn’t feel like our role is to tell chefs what they should cook or how they should cook it. We are not trying to replace their experience or knowledge. We are trying to create a dialogue so that an AI system could say to a chef, for instance: the raspberries you found in the market today, one molecular pairing theory says seaweed would be a good match for them; and based on what has traditionally been paired with raspberries in North America, these spicy, tangy ingredients could also go well; meanwhile, here are some ingredients that would pair well and are local to you; and these ingredients would pair well and are dairy-free; these ingredients are low in salt… etc. What do you think? What does your experience tell you to do with this information? Which ingredients will you select, and how will you bring them together?

GN: Were there unique challenges (or opportunities) when it comes to teaching an AI to work with flavor and taste, realms that straddle an interesting line between science and art?

Dr. Michael Spranger: One challenge is how subjective and the specific flavour is. Taste an apple, and you probably have a different perception of its flavour in your mind to me. But also, you have just one perception of one apple out of 7,500 apple varieties. 

It’s difficult for any system to account for an individual’s personal experiences of ‘apple’, and the specific flavour data of each of the 7,500 apple varieties (and the millions of other ingredients in the world) is not currently kept in one place. So, given this lack of clarity, how to make recommendations and suggestions surrounding flavour and taste? 

This is a huge challenge, but it is also what AI is particularly well suited to help with: personalization and accumulation are rich tools for us to explore.

GN: What have been the biggest surprises when it comes to flavors?

Dr. Michael Spranger: One surprise is how, for gourmet chefs, it is not only the concept of ‘this tastes good’ that matters in flavour. They also care about whether the flavour has a good story to it. Or whether the flavour is a new one. Or whether the flavour matches twelve other flavours they have on the menu. 

This presents another challenge for AI, which is to understand the motivation for a recipe… maybe you want to create something that has never been done before. Maybe you want to create something that has been done once before by one particularly famous chef. Maybe you want to create something that has been done for centuries, in a particular way, by a city of people. 

In terms of surprising flavour combinations, there are many! A personal favourite was chocolate, junmai sake, and cauliflower. 

Artificial Intelligence


Source: Robotics - zdnet.com

WA Auditor-General drags local governments over horrendous cyber risk management

Hit by ransomware? Make sure you don't make this first obvious mistake