FDA trains AI to sniff out bad seafood
The Food and Drug Administration is using artificial intelligence and machine learning to screen imported seafood for safety.
In a proof of concept, the FDA trained a screening tool to identify seafood that might be unsafe with years of historical data on seafood shipments that had been refused entry or required additional examination. According to a blog by FDA Commissioner Stephen Hahn, the test showed the technology’s ability to analyze data -- automatically identifying hidden connections and patterns that current screening was unable to catch -- could “almost triple” the likelihood that screeners will identify products with health and safety issues.
The second phase of the project, currently underway, calls for applying the machine learning algorithm in a field test of imported seafood. The plan is to use the model to help FDA staff decide which shipments to examine and what food in that shipment should be sent to a lab for testing. FDA will compare the model’s success rate with that of its current system in its evaluation.
Besides ensuring Americans have safe seafood options, the pilot also gives the agency a chance to learn how to leverage the data it already has from screening millions of import shipments each year and use it to power predictive analytics. The FDA collects tens of millions of data points on the producers, processors, exporters and sellers that ship food to the U.S., cross-checking it against its internal databases. One of the goals of the pilot, Hahn said, is to assess the ability of artificial intelligence and machine learning “to more quickly, efficiently, and comprehensively take advantage of all the data and information residing in our systems.” That insight can help the FDA decide which facilities to inspect, which foods are likely to sicken consumers and make other risk-based decisions.
“Imagine having a tool that expedites the clearance of legitimate, compliant shipments and improves by 300 percent our ability to know which shipping container to examine because that container is more likely to have violative products. It would save an immense amount of time, and potentially lives,“ Hahn said of the pilot. His remarks came on July 13, when the FDA announced its New Era of Smarter Food Safety Blueprint, a 10-year plan to leverage technology to improve food safety through the use of AI, the internet of things, sensor technologies and blockchain.
Connect with the GCN staff on Twitter @GCNtech.