By Artificial intelligence and the need for human direction was the topic of discussion at this year’s Lamont Rhodes Lecture at Northern State University on Tuesday, April 8.
“AI provides a different lens through which to see the world by sifting through a wealth of information,” said James Koltes, an associate professor at Iowa State University. “But having a human expert direct what should be done with the data is critical.”
The Lamont Rhodes Lecture Series is an annual talk made possible through support from the Rhodes and Lamont families.
MORE: Agriculture leadership class gains world perspective during trip to Argentina
Movies like “I, Robot”, and “The Terminator” share frightening scenarios of artificial intelligence going too far.
“We are not there yet, and hopefully never decide to try to go there, but people have that possibility in the
Koltes
back of their mind when they hear references to AI because of ideas spread by popular culture,” he said.
Koltes shared how AI is currently used in agriculture and business, the challenges encountered and the potential for the future.
His work in Iowa focuses on improving the health and efficiency of dairy cattle using genetics. He’s used tools like sensors and biomarkers to advance farming practices and is currently studying how non-coding DNA, once thought to be junk and without purpose, can apply to genetic selection. He uses AI to sift through the overwhelming amount of published genetic data to help with these efforts.
Koltes acknowledges that there is fear and mistrust surrounding AI, but many advances are important for society, especially in agriculture. Enormous amounts of data provide a way to look for patterns that can identify problems or solutions quickly. The quality of the data provided is essential. Those directing AI can use it to streamline time-consuming tasks carefully, focusing on what and why rather than just the ins and outs.
Focus on dairy animals
With a wry tone, Koltes acknowledged that the initials AI can lead to confusion as AI can refer to artificial intelligence, artificial insemination or Avian influenza. His message focused on the first.
In the cattle industry, he said it’s exciting to use AI to generate software to develop sensors that show when animals are healthy, when they are going through estrus (in heat) and when they are sick. The sensors can track feed intake, providing owners with numbers that might indicate when animals are ill or stressed.
Risks and rewards
The biggest challenge is unrealistic expectations, Koltes said.
At times, AI struggles with responses as it doesn’t know how to say “I don’t know” when faced with complex decisions based on complex rules, or it uses parroting to respond. Neither is helpful, he said. AI does very well with simple rules and repetitive tasks.
AI is rapidly advancing to improve its response to complex questions, but is not going to always give reliable responses when context is missing or human experience and intuition is needed.
“The goal is to advance complex reasoning,” Koltes said.
James Koltes, an associate professor at Iowa State University, discussed artificial intelligence in agriculture at this year’s Lamont Rhodes Lecture earlier this month on the campus of Northern State University. Northern State photo by Elizabeth Varin.
While there are rewards, there are also plenty of risks like keeping private data secure. There are ethical considerations, too, because the search for information can sometimes reveal too much.
MORE: Ellweins continue family farming tradition near Frederick
“The elephant in the room is who owns the data and why it is being generated. Data is valuable, and safeguards are needed,” Koltes said.
One of the most critical challenges with AI moving into the future is determining how to prioritize the path forward when priorities for research are often dictated by the funder’s interests.
Artificial intelligence can also help with data analysis to help find answers quicker. As an example, Koltes referred to a case from the mid-1930s when a farmer showed up at a lab at the University of Wisconsin in Madison.
The farmer was concerned about his recent cattle losses. He brought in some yellow sweet clover the cattle ate in the days leading up to excessive bleeding and death. Eventually, researchers were able to identify a mold on the hay that created a blood-thinning compound called dicoumarol.
Initially used as a rat poison, dicoumarol, also known as Coumadin, is now used as a medication to treat clotting disorders and prevent strokes.
“The concern for the health of the animal initially was the focus of the researchers at the university. They did not know that their research would lead to medical advances in the future. I hope that desire to help people will continue and that AI can help,” Koltes said.
Examples in agriculture
As AI evolves, agriculture will benefit from applications. Autonomous tractors will drive across fields, and drones can scout fields to identify potential threats from disease or insects. Some applications are already in use today for precision pesticide application.
By using satellite and drone images, there is an opportunity to predict what nitrogen levels are available in fields, what crop losses could be and potential yields and maturity levels. Koltes credited colleagues at Iowa State within the Translational Artificial Intelligence Center for developing and sharing those applications.
But, Koltes said, when looking for answers, it’s important to identify the correct data. For example, he said researchers tracked the movements of dairy cows, watching data points on images to predict animal lameness. Through their information, they predicted which cows would go lame with roughly 65% accuracy, he said.
As they looked over their results, one dairy expert reviewed what they were doing. He suggested that instead of watching the legs and gait of the animals, they should watch the arch of the cow’s back as an earlier indicator that she would go lame. With that data tracked, the accuracy was closer to 90%, Koltes said.
Predictions will not work without the right data, and regardless of the data gathered, a skilled expert must provide input, he said.
Take-home message
Koltes said AI is doing amazing things when it comes to image analysis and automated repetitive tasks. It is also advancing quickly, but is only as good as the data provided, and bad information can lead to mistakes.
While complex decision-making isn’t yet possible, he said, AI can be helpful, but it takes a knowledgeable person to evaluate how to apply that information.
MORE: Farmers welcome snow, rain as planting begins
“There is a lot of hype going around regarding AI, and we have much to learn,” Koltes said. “There are reasons to be excited, but make sure expectations for AI are realistic. People won’t be replaced by AI, but they will be replaced by someone who knows how to use AI.“
Connie Sieh Groop of Frederick is a veteran journalist and writer who has focused on all aspects of the agriculture industry during her career.
Connie Sieh Groop
Connie Sieh Groop of Frederick is a veteran journalist and writer who has focused on all aspects of the agriculture industry during her career.