Examining the potential of AI to drive nutritional innovation and increase productivity
In recent years, AI’s tremendous evolution enables new rates of innovation and productivity gains. The AI Institute for Next-Generation Food Systems (AIFS) at the University of California (UC) Davis, US, centers around AI to overcome key challenges in food production for sustainable nutrition.
Nutrition Insight discusses the potential of AI in nutrition industries with AIFS director Ilias Tagkopoulos, Ph.D., professor of Computer Science at the UC Davis Genome Center.
Tagkopoulos says AI is a great tool “to increase the productivity of what we are doing, doing more with less, and to accelerate innovation. Instead of a huge universe of things you can do, how can you make it less serendipitous, less by chance and more targeted in the discoveries of our world.”
“AI is amazing at taking a lot of complex data with complex relationships that humans alone cannot analyze and figuring out the associations across the different data points and variables, determining patterns and making recommendations — if you want this outcome, AI can be a tremendous help.”
He adds that AI is fast and efficient if users have the right computational power, algorithms and data. The technology is most impactful in repetitive tasks by reducing cost and time tremendously.

New type of AI
Tagkopoulos says that versions of AI have existed since the 1950s, though the technology is evolving substantially.
“In the past 10 to 15 years, we’ve seen a new type of AI emerge due to a substantial increase of training data, computational power and some new algorithms. This trifecta is the perfect storm, and we’re living in this renaissance regarding AI and what it can do.”
The technology has become a significant opportunity and concern, leading governments to collaborate and develop programs to manage, nurture and regulate AI.
“In the US, this was what gave birth to the AI institutes in 2020, through a competitive program that was managed by NSF (the National Science Foundation), together with other partners like the USDA (US Department of Agriculture),” explains Tagkopoulos.
These partners combined funding to create AI centers and institutes, like AIFS, which was established in October 2020.
Tagkopoulos says that AI can increase productivity, for example, in a factory or field.The institute aims to solve the world’s biggest challenges in crop and food production, ensuring a sustainable, nutritious, efficient and safe food supply while mitigating the impacts of climate change.
“We were given US$20 million to leverage AI to improve the food system. This collaboration is between UC Berkeley, UC Davis, the University of Illinois, Cornell University, UC ANR (UC Agriculture and Natural Resources) and the USDA. And, of course, we have dozens of partners, such as other universities, land grant institutions and other centers worldwide.”
“We look at the food system as a continuum, starting from molecular breeding, agricultural production, food processing and nutrition to the consumer,” explains Tagkopoulos. “It covers what you breed, the seeds you’re putting in, how you grow the plants, how you process it and give it to the market.”
Productivity and innovation gains
AI can be used throughout food systems to increase productivity and innovation.
Producers can reduce costs by increasing productivity in a factory or field, suggests Tagkopoulos. “For example, you’re using fewer ingredients when manufacturing your consumer-packaged goods or picking 10% more apples on the farm.”
AI will also convert faster innovative solutions. “Instead of doing 1,000 experiments to find the right one that will give you the molecule, now you only have to do a few iterations of a dozen of experiments.”
For example, AI can help determine what feedstock to use in food production and where because it has the right molecules and processing output.
“What will be the right formulation for a great snack or supplement for a specific target, such as anti-aging, anti-inflammatory or another bioactivity? AI can capture and formulate much more efficiently, with less cost, more impact and more activity,” adds Tagkopoulos.
AI can accelerate innovation, for example, helping to determine the formulation for an anti-aging supplement.“AI can help ensure you’re producing it faster, and it can set up the formulation, extrusion, extraction, fermentation and other parameters.”
Finally, AI can help target consumers, ensuring products are more tailored. “It connects the right consumer with the right product, so you maximize impact. To do that, it can take into account personal preferences, budgetary constraints and nutritional needs.”
“Building this internal knowledge also de-risks your business because if your staff leave the company, all their knowledge gets lost. With AI, this stays in the company and improves with time because it learns from data and experience.”
AI limitations
At the same time, AI is not without limitations. Tagkopoulos says that the technology, especially machine learning, needs data.
“Without data, AI cannot do anything. That’s very different from other algorithms that can capture the system without data because you’re programming them to do so.”
This means that it doesn’t work in solving data-poor challenges. “For example, if you want to figure out what feedstock to use if you don’t know its molecular composition, there is little AI can do to help you.”
He adds that AI cannot replace humans yet and needs to learn more about how the world works. “It still has biases because the training data we provide may be looking in one direction. We have seen that in healthcare — if your training data for an AI predictor has been based on white males, 30 to 45 years old, and you’re a woman of Asian or African American descent, at the age of 60, the system can’t perform well because of bias.”
Tagkopoulos stresses the importance of being aware of and accounting for these biases. “People are doing that a lot, but still, our world is very complex, and we need to do that more.”
Finally, he says that interacting with the physical world is difficult for AI. “We don’t have many robots going around seamlessly, being able to learn from the environment and evolve by themselves, beyond what they’re programmed into. Democratizing ways for AI to physically interact with the real environment will have transformative effects for our society.”