The Center for Data Innovation spoke with Davar Ardalan, founder of IVOW AI, a startup that is developing a culturally intelligent chatbot called Sina, and Alva Lim, co-founder of the Timor-Leste Food Lab, which is using Sina to preserve and promote traditional Timorese cuisine. The women discussed how to transform traditional knowledge into an AI-suitable dataset and how AI systems can support food innovation.
Hodan Omaar: Davar, what is the idea behind the work you do at IVOW?
Davar Ardalan: Our team at IVOW are journalists and developers. IVOW stands for Intelligent Voices of Wisdom, and we are developing a storytelling chatbot called Sina that has a unique appreciation for global cultures. As a storyteller, Sina can help preserve cultural knowledge and share those stories. For instance, we are designing her to learn about vibrant food cultures from around the world.
The idea for Sina was borne from my own experiences as a storyteller and from working with other storytellers. I spent many years at NPR News looking at how to reach diverse audiences through audio and later, I was managing editor at Hanson Robotics, which developed Sophia the robot. Each time Sophia travels to a new country—and she’s been to more than 30—she interacts with people from various backgrounds. That got me thinking about how technology shapes the way we consume stories, whether through print, radio, television, podcasts, and now AI!
Just as the printing press and the Internet revolutionized the way humans share information, AI tools are changing how people learn and interact with one another. Our goal is to build a tool that has a cultural richness to it, one that represents all the people that use it and that people use to preserve cultural and traditional knowledge. One of our first focuses is on food traditions from around the world.
Omaar: How do you get traditional knowledge, which exists for the most part in the form of oral histories and ancient texts, into a format an AI system can learn from?
Ardalan: The main idea is to start by having local users collect unique recipes and stories from different cultures. At the AI for Good Summit in 2020, we shared a vision for how one can translate recipes, such as the ones Alva Lim shared with us, through our Indigenous Knowledge Graph. This is our educational offering and is separate from the commercial application we are building focused on traditional foods from around the world.
In essence, a recipe represents a collection of distinct components including ingredients, instructions, techniques, and occasions, so we can format the recipe to separate it out into these components. We can then tag the data, meaning we add information to it or draw information from it. For instance, we can tag a recipe to indicate if it is modern, historical, or linked to folklore. We can also tag it with the Sustainable Development Goal (SDG) that it is most relevant to, such as SDG 3, which is good health and well-being. Tagging is important because it gives Sina additional information that may impact how she parses information and where she might look to for answers to questions. The final result is an ontology that sorts information into logical hierarchical relationships, which we use to train Sina.
Omaar: Alva, how are you using Sina to preserve knowledge about traditional food?
Alva Lim: At the Timor-Leste Food Lab, our team of cooks, chefs, food storytellers, and local food champions are working with IVOW to help preserve and promote traditional knowledge about Timorese cuisine and native ingredients like wild yams, beans, and vines.
There are a few challenges with preserving traditional food knowledge. First, there are issues of authority with getting the knowledge and sharing it. For example, when some members of our team were going home to their villages over the summer, I wanted them to ask their older relatives about different recipes they use. But my colleagues were hesitant because there are certain rules about who can ask such questions, how they can ask, and who has the permission to tell. That makes conserving traditional knowledge more difficult. The second challenge is that the food habits of younger generations are changing and for some, they are moving away from wanting to eat traditional food because of the negative framing that can surround it. For instance, some people look down on traditional dishes like batar da’an, which is a kind of corn stew, as “poor people’s food.”
By partnering with IVOW, we want to use AI systems to not only preserve images, videos, and recipes of traditional food but to ensure that this knowledge is heard. This will help restore confidence and pride in our indigenous foods but also conserve knowledge on the complexities of preparing and eating them: Knowing when to look for specific ingredients, how to harvest them, the techniques for preparing different foods, and the cultural rules on how you can eat them. Conversational AI tools like Sina can also really help appeal to the younger generation. More than 70 percent of the population in Timor is under 30 years old.
Omaar: What about food innovation? Can AI help with that?
Lim: Oh, yes! A big part of what we do at the lab is food innovation. There is such a diverse range of indigenous vegetables, fruits, tubers, grains, and wild edibles in Timor that have nutritional value but are being forgotten about and replaced by crops like white rice. AI can support food storytelling and food innovation to promote better livelihoods. For instance, there is a small island in Timor called Atauro Island where people have long survived off the land because they learned how to grow and eat a plant called moringa. Back in the day, they didn’t have nutritional science but they understood the value of moringa from the way eating it made them feel. That’s why it became a staple for them and no dish was complete without it. Today, we’re trying to use AI and food storytelling to encourage people to expand the range of foods they’re eating and draw on traditional knowledge.
What we are trying to do with food is somewhat similar to what Japanese people have done with architecture. Consider that for more than 1,000 years, local people in Japan have been tearing down a holy shrine called Ise Jingu every few decades, only to rebuild it anew. The idea is to pass on valuable building techniques from generation to generation while still evolving. In this way, modern architectural practices are connected to the old but technology has allowed the building itself to adapt over time. Using technologies like AI, we too can co-create new food combinations and traditions.
Omaar: Looking forward, what’s next for IVOW and Timor-Leste Food Lab?
Ardalan: We are currently in product development. As a culturally oriented application dedicated to preserving traditional dishes and historical stories around them, we will use AI tools to recreate lost dishes and forgotten food traditions. Our Sina AI features will help users recreate traditional dishes based on recipes and stories right from the kitchens of users from around the world.
Lim: In the future, I hope to see AI-powered tools like Sina integrating with existing geospatial technologies and applications like Google Earth to give users a truly immersive experience. Imagine if a user could look at a map and see where an ingredient, plant, or woven fabric comes from and ask Sina to provide more information about it. Developments in natural language processing could help Sina provide varied information on everything from recipes, to harvest methods, to nutritional analysis in more languages, including indigenous ones.