Obtain the data you need to make the most informed decisions by accessing our extensive portfolio of information, analytics, and expertise. Sign in to the product or service center of your choice.
Article: Experts offer tips on AI in food and agriculture at IFT meeting
29 July 2020
From helping farmers track animal diseases to predicting the
likelihood of pathogens in fields and water, artificial
intelligence (AI) has a vast potential for application throughout
the food industry.
If applied correctly, technologies such as AI, machine learning
and predictive analytics can help the food industry boost
efficiency, reduce food waste, improve food safety and support
sustainability, said top-level food industry experts who discussed
the applications of AI in the food supply chain at the Institute of
Food Technologists' (IFT) virtual conference SHIFT20. .
However, there are still many challenges the food system has to
overcome before expanding the use of the novel technology, such as
creating basic infrastructure, improving the quality of data and
bridging a gap in the availability of technical experts who can
help food firms integrate AI into their operations.
Food companies should also be aware that AI is not a silver
bullet, and it may not always be the right technology for the job,
said experts.
"In some cases, it's a great tool to have, in other cases there
are better, more appropriate tools out there to solve some of those
issues," said Martin Wiedmann, a leading food science expert and
the Gellert Family professor in food safety at Cornell
University.
During a "virtual fireside chat" event, Wiedmann and Simon Liu,
associate administrator at USDA's Agricultural Research Service
(ARS), mapped out the current and future uses of novel technologies
such as AI, predictive analytics and machine learning throughout
the food supply chain.
They also provided practical tips for companies who want to test
ways to optimize their operations using AI and machine
learning.
Experts outline steps for application of AI, machine
learning
Since data is key for AI or machine learning to work, companies
should first ensure they are collecting the right, quality data and
establishing a good basic data infrastructure, Wiedmann said. That
is very important, especially in food industry sectors such as
produce, which has only recently started converting from paper to
digital records.
In such a conversion, small issues can make a big difference and
having a database in which the word "cantaloupe" is spelled five
different ways is not going to be useful, Wiedmann noted.
Companies should also make sure they are asking the right
question, to ensure that the machine learning or AI technology can
provide an analysis that is useful and practical for the company.
Companies should always complete a pilot first to ensure not only
the technology works in the way they want, but also that AI or
machine learning is the right technology for the goals they are
trying to achieve, Wiedmann advised.
"In some problems, you can use mechanistic models, models that
use known mechanisms to predict and they tend to be much better at
predictions," he noted. "I am continuously amazed by the
underutilization of very standardized tools such as what we call
simulations in the food industry. They are not difficult to build
and are very powerful."
Therefore, companies should first do a pilot focusing on
"something you are willing to change," Wiedmann advised.
"And if your AI predictions aren't right - and often they will
not be - then figure out why and learn from that and continue to
improve."
USDA's ARS explores uses for AI application
When applied correctly, AI can have tremendous potential to
improve the food system, and scientists at educational institutions
and USDA are pushing to expand the areas where the new technology
can improve the efficiency of the food system.
According to Liu, USDA already has a number of initiatives
underway that are exploring the uses of AI in the food and
agriculture sectors.
"At ARS we have initiated a project we call the 'big data'
project," he said. "This is the sixth year of this big data
project, so right now we have many different projects around the
country, we have many uses around the country and we are continuing
to expand this now."
Currently, the most common uses of AI and machine learning focus
on issues related to food production, such as improving crop
rotation, streamlining crop planning, and monitoring crop and soil
health or deficiencies, Liu said.
For instance, researchers at USDA's Beltsville Agricultural
Research Center in Maryland are piloting a low-cost automated
detection system, which uses computer vision coupled with machine
algorithm to document conditions and problems in corn and soybean
crops, Liu said.
With the help of Microsoft, the project allows researchers to
gather data from multiple sources, including sensors, drones,
tractors and satellites, feed it into cloud-based artificial
intelligence models, resulting in a detailed picture of conditions
on the experimental farms. While still at an early stage,
researchers want to use the program to develop web-based tools to
help farmers allocate resources more efficiently.
In addition, ARS is exploring the use of AI in disease detection
- a major threat to the economy and food security, Liu said.
"AI-based image recognition systems could recognize specific
plant diseases with a high degree of accuracy and potentially pave
the way for field-based crop disease identification using mobile
devices, such as smartphones," he said.
ARS scientists in New Mexico, for example, have developed
AI-based tools to provide early warning of pest and disease
outbreaks.
Researchers are also looking to use predictive analytics to
forecast effective crop production and yield and determine how much
of the crop will be harvested under specific conditions. Scientists
are testing that approach in Iowa by using the new technology to
determine potential strategies to increase profitability of each
field based on varied uses of nutrients and pesticides, Liu
said.
Novel technologies applied further down the supply chain can
improve quality and boost food safety, suggested Wiedmann.
Image recognition, for example, can ensure product quality and
track defects in food production facilities, Wiedmann said. Facial
recognition, on the other hand, can determine whether food facility
staff is following safety procedures, such as washing hands, he
noted.
Visual imaging with AI has a potential for application in animal
slaughter and meat processing, particularly when it comes to less
homogenous cuts of meat, such as pork halves. To use a more
automated system to cut such products, you can take images, let AI
analyze them and then guide knives as they cut the product,
Wiedmann said.
It is important to remember that AI "really shines" in areas
where large data sets are already available - whether those are
GIS, meteorological or supply chain data, he stressed.
Future AI applications in food regulation, distribution,
marketing
AI can also have useful applications for regulators, who can use
data on companies' history of inspection to determine frequency and
priority of inspections, Wiedmann noted.
Future applications of AI could help significantly improve
sustainability and reduce waste in the food industry. For example,
it could predict when people are going to buy certain products and
prevent an oversupply or undersupply of food.
AI has the power to improve shelf-life labeling, and scientists
at Cornell are already testing ways to use the technology to make
"sell by" or "use by" dates, Wiedmann noted.
"Just more precision to some of these 'fudge factors' I am going
to call them that [industry uses] right now will reduce food waste
and have a huge impact on sustainability," he said.
As AI and machine learning technologies continue to expand,
there are also areas at the end of the supply chain that can
benefit - including the food distribution and online shopping
sectors.
Using an AI ordering algorithm, for instance, can help online
grocery stores - which are becoming increasingly popular - increase
efficiency and reduce food waste, Liu noted. The new technology
could help consumers improve what they eat, as imaging recognition
applications could estimate the nutrition content of various foods,
he added.
"Maybe now we are helping folks in the front of the food supply
chain, but in the future, we will be moving to cover the back end -
the distribution, the food marketing and waste management," Liu
said.
Another frontier for the future of AI would be taking the
technology from "predictive" to "prescriptive," and using it not
just to predict when issues may arise but also how to address them,
Liu noted.
Challenges for expansion
Yet, a number of challenges still exist before the food industry
can harness the full power of technologies such as AI, machine
learning and predictive analytics, experts stressed.
One big issue for food companies now is the gap in skilled
experts who can help companies adopt AI and machine learning
technologies into their operations. That is a significant challenge
and it can only be addressed using a multi-prong approach that
requires the public, private and academic sectors to work to
address collaboratively, Liu noted.
"We need to have a talent pool, to make sure we have the people,
the staff that have enough management skill or business skill to
advance the AI application," he said.
Building appropriate technical and data infrastructure can also
be a challenge for companies as not all food firms can afford to
divert resources to such initiatives, the experts noted.
USDA is also trying to figure out how to bring together the AI
resources from different locations around the country, Liu
noted.
"How can we structure that expertise so that we can accumulate
the AI know-how and then be able to leverage that expertise
properly," he stressed.