Researchers look to AI for more personalized food
A UC Davis-led institute is building open agricultural datasets to help machine learning connect farming, food and health.
By Daniel Okafor · Business Editor
3 min read
Artificial intelligence could eventually help produce foods designed for the nutritional needs of specific groups, according to Ilias Tagkopoulos, a computer science professor at the University of California at Davis. The idea matters because researchers say food and agriculture still lack the shared data needed to use machine learning at the scale seen in other fields.
Tagkopoulos told Fortune that deep learning, which detects patterns across large datasets, could reveal links between diet and health that researchers have not yet identified. One possible outcome, he said, could be apples or other foods engineered with nutrients matched to the people most likely to eat them.
The work is part of the AI Institute for Food Systems, known as AIFS, where Tagkopoulos serves as director and principal investigator. Fortune reported that the government-funded project began last year to study how machine learning could improve farming and food distribution.
The data gap in agriculture
Tagkopoulos said artificial intelligence has produced stronger software in some industries but has had less effect on agriculture. He pointed to a shortage of freely available food and farming data for training machine-learning systems.
AIFS aims to build and maintain an agricultural dataset modeled on ImageNet, the large labeled image collection that helped advance computer-vision tools. Fortune reported that the food-systems version would include annotated photos of crop fields and sensor readings from internet-connected farm devices.
Those devices could include tools such as thermometers that farmers use to track air temperature and humidity. AIFS plans to clean, label and publish the data for free use by others, according to Fortune.
The institute also includes researchers from Cornell University, the University of Illinois and other institutions. Beyond data collection, AIFS wants to link artificial-intelligence researchers with entrepreneurs and experts across the food supply chain, Fortune reported.
Tagkopoulos said venture capital investors have paid less attention to food and agriculture than to areas such as health care and enterprise software. He said AIFS could help bring separate groups together, with open agricultural data serving as a starting point.
A long path to tailored nutrition
The institute’s broader goal is to support research that could lead to genetically modified foods with nutrients and properties aimed at particular populations, according to Fortune. Tagkopoulos gave diabetes as an example, saying people with the condition could one day buy foods that are better suited to them at a molecular level than current options.
That outcome remains far off. Tagkopoulos said such a system is years away, and possibly decades away, because researchers would need far more data to build and test the required artificial-intelligence models.
For now, the work centers on creating the basic infrastructure that machine learning needs: reliable, organized and accessible information from farms and food systems. Tagkopoulos described data as essential fuel for modern AI software, and AIFS is trying to supply more of it to a sector where he says it has been scarce.
This story draws on original reporting from Fortune.