How Artificial Intelligence in Food Safety Will Affect Food Processors
Dr. Ben Miller
There’s no doubt that artificial intelligence (AI) is reshaping the world as we know it. But what is AI, really, and how will artificial intelligence in food safety evolve?
In our webinar, The Use of Artificial Intelligence (AI) in Food Safety and What to Expect Next, Dr. Ben Miller, VP of Regulatory and Scientific Affairs at The Acheson Group (TAG), led a fascinating discussion on how the food industry is adapting alongside this new wave of technology.
In order to stay competitive, food processors of all industries must understand the fundamentals of what artificial intelligence in food safety is, and how it will continue to reshape the industry in the years to come.
What is AI?
Artificial intelligence is a field that combines computer science and datasets to solve problems. There are two types of AI: Strong AI and Weak AI.
Weak AI, or Artificial Narrow Intelligence (ANI), is software that is trained to solve a “narrow,” specific scope of problems. Siri, Alexa, and Chat GPT are all types of ANI.
Strong AI is the theoretical concept that keeps Hollywood busy with its ever-expanding repository of sci-fi thrillers: a conscious artificial intelligence with capabilities equal to, or even exceeding, that of human intelligence. To date, Strong AI has not yet been created. (Famous last words.)
Artificial Intelligence vs Machine Learning vs Deep Learning — What’s the difference?
Today’s computer science terminology can cause some confusion when discussing AI — but there’s no mystery involved and it’s easy to understand once you know the basics.
Nested within the all-encompassing field of AI are the sub-categories of machine learning, neural networks, and deep learning.
Machine Learning
Machine learning builds on AI by continuously using data to increase its problem-solving accuracy.
Neural Networks
Neural networks are based on the concept of generating artificial neurons (computer nodes) that simulate the way the human brain recognizes patterns to solve problems. Without diving too much into the details, data is quickly passed between node layers only if certain conditions are met to reach a conclusion. Neural networks utilize training data (machine learning) to improve their ability to solve problems (AI) over time.
Deep Learning
Finally, a neural network with more than three node layers is described as deep learning.
The Risks of AI
As with most things in life, the biggest risk of artificial intelligence in food safety is in using a tool you don’t fully understand.
One of the biggest problems with AI is that it doesn’t “show its work,” proof of the sources, and information it uses to draw conclusions. These sources may have bias, mistakes, ambiguity, or other factors that lead to an incorrect conclusion. This is known as the prolific “black box” of AI: a user submits a query, the software completes millions of processes it trained itself how to do based on volumes of training datasets, and out comes a solution – not necessarily the solution, or even a correct solution.
These are not truth machines, they’re word association machines.
This is one of several major reasons why there is so much controversy surrounding AI. The only questions that should be asked of AI are ones that can be easily verified for accuracy either through the knowledge of a subject matter expert or through reverse engineering the data the software was fed to obtain the same conclusion.
Current and future food safety AI applications
When it comes to artificial intelligence in food safety, there are a number of ways to use AI to improve food safety without causing unnecessary risks to consumers or food processors. The key is in applying the technology to easily verifiable tasks.
The best general use cases for getting started with artificial intelligence in food safety include:
brainstorming
idea generation
and outlining procedures
It’s easy to see the potential for artificial intelligence in food safety. Supply chains can utilize environmental data to make predictions about areas that might be at higher risk for pathogens. Nontraditional data sources — such as social media or online reviews — combined with natural language processing models could be used to detect foodborne outbreaks. Between computer vision, natural language processing, and analytical tools, there are a wide variety of food safety applications for AI.
Notable uses of artificial intelligence in food safety
Optical imaging with AI has been used to detect E.coli significantly faster than in traditional approaches and opens the door to potentially automating ongoing bacterial detection processes.
The FDA Artificial Intelligence (AI) Imported Seafood Pilot program utilizes machine learning to predict the risk of and identify contaminated imported seafood products that may pose a public health hazard.
Western Growers in partnership with Creme Global uses ML trained on data from inspections, product and water testing, location, meteorological, and landscape sources to predict food safety issues associated with produce.
AI models utilizing behavioral data from workers in food processing facilities — such as handwashing, food handling processes, traffic flow, training performance, turnover data, workforce tenure, and absenteeism — can identify patterns to evaluate the level of risk.
As technology evolves, more regulations will follow. Businesses must stay up-to-date with the latest compliance and FSMA requirements to ensure their processes and tech align with the latest regulations.
Prepare for a new wave of digital transformation
You can’t train good AI without good data. A major problem with only keeping hard copies of food safety data, such as critical control points or preventive controls, is that there is limited data standardization between companies. The more standardized the data, the more training data sets the machine learning has to improve its accuracy. A collaborative shift towards digitization will empower the industry to integrate advanced artificial intelligence in food safety.
Furthermore, if an open-source model applicable to your business was developed, your data would need to be ready to go in a digitized format for your company to reap the benefits.
The solution is for companies to start digitizing paper-based processes now. Making small, incremental technology investments based on global predictions will prevent major lump-sum expenses down the line when competitors have already prepared to implement AI food safety technology into their daily processes.
About the author: Dr. Ben Miller has nearly 20 years of experience in food safety regulation, epidemiology, outbreak investigation and response, and public health. Ben's previous work includes serving as the Division Director of the Food and Feed Safety Division at the Minnesota Department of Agriculture. He's also held leadership roles within national regulatory associations including serving as Director for the Association of Food and Drug Officials, as chair of the FDA’s Manufactured Food Regulatory Program Alliance, and co-chair of the FDA’s Partnership for Food Protection’s Response and Recall workgroup.
