UC Davis Institute targets next-gen food systems with AI-powered nutrition
Recognizing AI’s potential to create healthier and more sustainable nutrition, the AI Institute for Next-Generation Food Systems (AIFS) aims to transform the food system from developing and growing new varieties to food production and nutrition. The US Department of Agriculture-backed AI Institute, headquartered at the University of California (UC), Davis, views AI as the “connective tissue that brings this all together.”
Ilias Tagkopoulos, Ph.D., AIFS director, tells Nutrition Insight that the multi-trillion-dollar food industry is ripe for disruption. “We have mapped only a small percentage of the universe of chemicals in our food, and there is so much to learn on what they do to our health.”
“We have done a lot of studies, but there is much more to learn given the vast complexity of our physiology and the food matrix. With Ozempic and similar drugs, we have seen how one peptide can change the world. We need better nutrition, information, and lifestyles that will lead to a better quality of life for all of us, ensuring that we keep being healthy by preventing disease and not just treating it after we get sick,” says Tagkopoulos.
“Nutrition is a key piece of the transition from sick care to healthcare. To achieve that, we need to design, grow, process, and consume food that is good for us and the planet.”
The institute brings together researchers from six institutions. It funds and conducts research, education, and outreach with its partners and collaborates with the industry to offer access to AI research. Tagkopoulos emphasizes that the institute strongly encourages interdisciplinary and multi-institutional requests for proposals.
Nutrition investments
Tagkopoulos sees great potential through AI in developing foods that are easy to make, inexpensive, healthy, and taste great. For example, the Davis-born company Rivalz Snacks developed healthy snacks with AI-powered and nutrient-rich indulgence.
However, he says these examples are limited. “Why don’t we have more of that? The answer is that the investment, especially in AI for food, is still lagging other sectors, such as pharma.”
Although some visionary food companies invest in AI, Tagkopoulos says it is challenging to attract AI talent for the food industry, which requires substantial investments. “If my students go to Google or the pharma industry, they will be paid more than twice what they would be paid in the agricultural and food industry.”
He says that at some point, investments in AI and technology will come to the food industry. “Until that happens, we will still struggle to go to that much-needed speed for delivering those products.”
Tagkopoulos says it is challenging to attract AI talent for the food industry, which requires substantial investments.Building a healthier world
Ensuring people eat healthier requires actions across a multitude of dimensions. First, you need to know what people eat and do “in a way that is frictionless and doesn’t take energy,” according to Tagkopoulos. “If you need to spend five minutes logging your meal or exercise every time you eat, maybe you won’t do it.”
“The next step is to know how your eating affects your health,” he continues. Tagkopoulos is part of a team that includes the Rockefeller Foundation, the American Heart Association, and dozens of researchers in various institutions that analyze the molecular composition of foods and then connect these molecules to bioactivities and health effects. The Periodic Table of Food Initiative (PTFI) has mapped over 500 foods and counting.
AIFS has also developed FoodAtlas, an evidence-based AI-powered tool that reads all available published literature and maps the relationships between food, chemicals, and diseases.
“FoodAtlas looks at the millions of papers published and builds a map of what foods are connected with what chemicals and at what quantities, and with which health effects,” Tagkopoulos explains. “Taste and texture are also very important when it comes to food, and this is another dimension where we see our AI resources expanding.”
“But how do you get it in your hands? One example is creating a mobile app. For example, when consumers take a picture of their food, connect it to FoodAtlas, which can tell what is in that food, what chemicals it contains, and their associated health benefits to determine its health.”
“Or we can have a dietitian AI that guides you on what to eat next and tells you whether you had a good day of eating or need more protein. This interaction is more natural than reading a paper or going to a database.”
Expanding impact
AIFS aims to expand its impact on people and the planet, creating a healthier society and a more sustainable planet. Delivering that impact by producing more nutritious and sustainable plants or food involves many aspects.
“It needs to be something that has the right molecules, that are healthy for us, and that has the right taste and appearance, so the sensory output that the consumer likes; otherwise, very few will buy it,” says Tagkopoulos.
Ensuring people eat healthier starts with knowing what they eat and how that affects their health.He adds that the same principle holds for price and product access — if products are too expensive or their shelf life is too limited, people won’t be able to buy them. “It has to be accessible to the consumers, either by offering it at a low price point or being subsidized by health insurance or other similar approaches to reach more people, regardless of age, geography or financial status.”
“For the planet, how do you ensure that cultivation or food production does not use many resources? How can you minimize the resources we use, such as land, irrigation and water, pesticides or fertilizers, to have a great product?”
The institute connects all these different aspects in its various projects. “We don’t think impact can be achieved significantly without collaboration and integration,” Tagkopoulos adds.
Developing new foods
Creating healthier and more sustainable varieties involves different phases in the food chain, from creating a variety to growing and processing it until it reaches a final consumer.
Tagkopoulos explains that such projects may start with questions like: “How do you use AI to have better genetics, use less water, be more drug-resistant, use less pesticides, but have the right molecules for a great taste?”
“Then you go to agricultural production — I have these amazing things, but how do I plant them? When do I plant them? When do I harvest them? How can I do this more effectively?”
Food production then looks at how to process these new foods — for example, maximizing the beneficial nutritional compounds in a finished product or reducing waste and adding value to byproducts to produce more with the same resources.
AI can also help develop foods that maximize the beneficial nutritional compounds or reduce waste and resource use.In the case of consumer-packaged goods, companies can use these ingredients in their formulations and processes to create healthy, tasty, and affordable food products. However, cautions Tagkopoulos, getting those recipes and processing equipment right can be a challenge.
“How do you determine the right values for your process parameters, regardless of whether that is baking, fermentation, or extrusion? These can easily be millions of combinations, which is impossible to test in a real scenario. This is where AI and other computational methods, such as computational chemistry, and fluid dynamics can help by reducing the number of experiments you have to do — dozens instead of millions. Ultimately, it is about being more productive and innovative, doing things faster, better, and cheaper.”
Tagkopoulos points out that the era of domain-specific large language models and specialized AI agents has just started to impact the multi-trillion-dollar food industry.
He says that researchers like him expect a combination of large-scale datasets, powerful AI algorithms, and a need for healthier products to transform the field with far-reaching and extensive implications for people’s health, lifestyle, and wallets.
