When artificial intelligence (AI) pioneer OpenAI unveiled ChatGPT in November 2022, prognosticators predicted that the revolutionary AI chatbot would swiftly usher in a new world order.
Algorithmic excellence, they prophesied, would fast-track solutions to everything from energy crises and housing shortages to climate change and cancer. But so far, progress has been slow. Research behemoth RAND says more than 80 percent of AI projects fail, and business technology consultant Gartner expects at least 30 percent of generative AI projects to be abandoned after the proof-of-concept stage by the end of 2025.
Clearly, AI hasn’t yet saved the world. But there are promising signs that it might soon be able to help feed it. And not a moment too soon, according to Vikram Adve, a computer science professor at the University of Illinois at Urbana-Champaign, who also serves as co-founder and co-director of the Center for Digital Agriculture and leader of the Artificial Intelligence for Future Agricultural Resilience, Management and Sustainability (AIFARMS) Institute.
“The world population is growing dramatically and will be over 9.5 billion by 2050,” Adve says. “That means we have to dramatically expand the world food supply, and we have to do it at the same time that the availability of land, fresh water and farm labor are decreasing. Essentially, we have fewer resources to produce (the food), but we have to produce a lot more.”
AI is ideally positioned to fill the growing gap between supply and demand by ingesting reams of data from satellites, drones and ground-based sensors, then expertly weaving it into practical intelligence that farmers can use when they’re making decisions about planting, tending and harvesting their crops.
“We have a deluge of data from novel sensing and computing technologies, but moving from data to actionable insights requires a paradigm shift,” says Baskar Ganapathysubramanian, a professor of mechanical engineering at Iowa State University, where he is principal investigator and director of the AI Institute for Resilient Agriculture (AIIRA). “AI provides that paradigm shift.”
On-Demand Expertise
One of the AIIRA projects Ganapathysubramanian is most excited about is the development of digital twins — virtual replicas of plants that can be manipulated and analyzed with AI to see how they’ll perform in specific conditions.
“If I have a digital twin of a plant on a computer, I can simulate a lot of ‘what if’ scenarios,” Ganapathysubramanian says. “What if the temperature is changing? What if I have more than 20 days without rain? What if I pack the plants together at a higher or lower density? To find the answers would take me four or five months of a growing season in the real world. But using a digital twin on a computer, it will take me maybe a day or two to simulate.”
With the results of this simulation capability, farmers would be able to predict which specific plant varieties will perform best in their fields and quickly determine the best solutions when faced with challenges such as extreme weather.
Another project AIIRA is working on is the development of an AI chatbot for pest management.
“We have collected millions of images of agriculturally relevant insects and weeds across the globe, and we are training a machine-learning model to identify them,” explains Ganapathysubramanian, who says farmers can take pictures and upload them to an open-source website capable of pinpointing the pest. Eventually, he notes, that same website will be able to advise farmers on what actions they should take to mitigate the pest.
AIFARMS is also working on an agricultural AI chatbot. “All farmers need advice,” says Adve, whose research team is training an AI chatbot with more than 400,000 technical documents about agricultural production so that farmers can ask questions and quickly get expert answers.
That’s exciting to growers such as Heather Hampton Knodle, who farms corn, soybeans and wheat for feedstocks in Fillmore, Ill.
“Production agriculture is a very competitive business, so what we really need is information,” says Knodle, who’s especially interested in AI-based insights about soil composition that could help her choose the ideal seeds to plant. “People like me who’ve been at this for a while are ready for some decision-making aids to help us take all of the data that we’re able to gather on our farms, ask advanced questions about it, and do analysis of varied scenarios to help us play to our strengths and size up exactly what our comparative and competitive advantages are.”
Crop science giant Bayer is piloting a generative-AI system to help its customer service reps provide such expertise. Trained with years of internal data, insights from thousands of trials within its vast testing network and centuries of aggregated experience from Bayer agronomists around the world, it can quickly and accurately answer questions related to agronomy, farm management and Bayer agricultural products, according to Amanda McClerren, CIO and head of digital transformation and information technology for Bayer’s Crop Science division.
“We’re focused on getting this tool in the hands of our customer-facing agronomic service employees to allow them to better serve our customers, spend more time on valuable questions with our customers and, honestly, connect with more customers,” McClerren says.
Unleashing Efficiency
Some of the most promising uses for AI in farming are related to “precision agriculture” — data-based farming that tailors crop management to individual fields and plants.
“Precision technologies … help farmers plant more precisely, manage each plant more effectively and harvest crops more efficiently,” says Sarah Schinckel, director of emerging technologies and architecture at John Deere. “AI is a key tool that powers many precision technologies that farmers are using in their fields and in their offices.”
An example is John Deere’s See & Spray system. The technology “uses computer vision and machine learning to detect where weeds are located in a field and apply herbicide only and exactly where it’s needed,” Schinckel says.
During the 2024 growing season, See & Spray was used on more than 1 million acres and saved farmers an estimated 8 million gallons of herbicide mix, Schinckel says.
John Deere also uses AI in its latest self-driving tractors. This enables soil tillage without an operator in the cab, freeing them “to handle other value-added tasks around the farm,” she says.
Indeed, Adve says, farms that struggle to find labor can use AI to make the most of a limited workforce. Take livestock, for example. “You can use cameras to monitor animals 24 hours a day, seven days a week, and (can) use AI to recognize early symptoms of heat stress and disease. When there’s an issue, human managers can be notified to come and deal with it,” he says.
That could be critical for producers such as Andrew Nelson, a software engineer and fifth-generation farmer who grows wheat and legumes in Farmington, Wash. “It’s getting increasingly hard to find skilled workers who want to work on farms,” says Nelson, whose farm is currently testing AI-powered cameras that monitor crops to detect problems such as water pooling, which can easily destroy plants if left to fester.
Brisk Breeding
Farmers who can’t afford cutting-edge equipment — adding John Deere’s See & Spray technology costs approximately $25,000, for example — can still benefit from AI thanks to upstream innovations in the supply chain. Bayer, for instance, is using AI to fine tune the seeds it develops.
“We have many, many years of data about all the (seed) lines in our pipeline — the grandparents, the parents, the progeny. We have all this information about their genetic material and how they perform in the field,” McClerren says. “A few years ago, our data scientists were able to develop an algorithm that can accurately predict the field performance of new seed based solely off that genetic information.”
With that algorithm, Bayer can create digital twins and simulate how different seeds will perform in various conditions, thereby hastening the development of crops with increased yields, drought resistance and other desirable traits.
“A traditionally bred crop takes 12 years to get to market … . Our process, from discovery to market, is three years,” says Brendan Collins, co-founder and CEO of Avalo, an AI startup that is similarly focused on accelerated crop breeding. Among the products Avalo is developing with AI-assisted breeding are a fast-growing broccoli variety that matures in 37 days, a heat-resistant tomato plant that retains its fruit in high temperatures and a resilient strain of cotton that requires less fertilizer and fewer pesticides.
“There is no single gene that makes your corn big or your strawberries red and juicy. It’s a process of thousands of genes interacting in really complex ways with each other and with a really dynamic, complicated environment,” Collins says. “Using AI, we can understand the genetic basis for the traits we care about and cross-plant parents in the right ways to get the outcomes we want.”
Higher Profits, Better Food
Although there are significant barriers to reckon with — the high cost of hardware and gaps in rural high-speed internet connectivity are two of the biggest — AI’s potential benefits are clear.
“It’s almost impossible for farmers to make a living these days,” Collins says, referring to rising input costs and falling commodity prices. “What AI is offering farmers is a way to use fewer inputs in a more explainable system that guarantees them more profit at the end of the day.”
And what’s good for farmers ultimately is good for consumers. “AI can accelerate our path toward using less chemicals, growing more resilient crops and producing higher-quality food,” Nelson says. “We need to use less and produce more, and to make more nutritious food. So, I’m pretty excited about that.”
发表回复