Will Artificial Intelligence Ever Be Smart Enough for Smallholder Farmers?

Dr. Neema Mduma, a Grow Further research partner at the Nelson Mandela African Institution of Science and Technology (NM-AIST), says the key to artificial intelligence technology is software capable of learning and recognizing patterns.

“When you are talking about artificial intelligence, inside artificial intelligence there is machine learning, and for machine learning, there are different functionalities,” she explained. Her work is “focusing on using machine learning and using algorithms focused on prediction.”

Grow Further is currently funding her project to develop an AI-driven smartphone app to identify crop diseases early when they’re easier to control. Unlike some AI applications that spit out information that can be false or misleading, her software recognizes its own limitations and gives farmers a probability that a particular disease is present along with recommendations of what to do if it is.

Recently, no technology has been in the news more than artificial intelligence, or AI for short. While its overall impact on society so far hasn’t matched most of the hype, experts agree that it’s only a matter of time before this changes.

Optimists say AI will solve a host of human-caused problems, including climate change. Pessimists say this technology will make millions of workers obsolete. Industrialized farms are already adopting these technologies, but if AI is to help us win the fight for better food security it will need to reach smallholder farmers.

Grow Further wants to know if AI can be unlocked to help smallholder farmers grow more food and earn better incomes. Dr. Mduma and her colleagues at NM-AIST are strongly convinced that the answer is “yes, it can.” For others, the answer could be “maybe, but,” meaning they think it will likely take further innovation and experimentation before smallholder farmers come to even know what AI is, let alone how it could be useful to them.

What is AI?

To many, the phrase “artificial intelligence” used to mean something very specific: A computer capable of thinking like a human, one even capable of self-awareness. This is the stuff of science fiction; machines smart enough to turn on their human creators as in the television series Battlestar Galactica or The Terminator film franchise.

Others have wildly expanded the definition of AI to mean any really fast, really smart computer. Many tech companies are now bragging that their search engines and social media sites are driven by AI only because the servers that support them can process huge volumes of data very quickly.

True artificial intelligence technology, the kind popularized by new chatbots, lies somewhere between these extremes.

In a summary of AI’s potential role in agricultural extension, Indian agriculture scientists Debi Jayasingh, Ashish Anand, and Kiran Das explain that AI is essentially computer software that’s been cleverly trained to detect patterns and make predictions. This is what Dr. Mduma and her team in Tanzania are working on, and it’s why they have to first “teach” the AI to recognize what plant diseases look like.

The AI chatbots millions of people are now becoming familiar with are nothing more than software programs trained to recognize patterns in human speech and written communication. They then mimic human communication by relying on the vast trove of data available on the internet. It works largely the same for auditory and visual processing. Thus, NM-AIST has to first upload thousands of images of crops plagued with pests and diseases to tell the AI system what to look for.

“A chatbot exposed to text examples can acquire the ability to generate natural-sounding conversations with individuals, while an image recognition tool can develop the capability to identify and describe objects within an image,” Jayasingh, Anand, and Das explain. They say AI isn’t so much software that thinks like a human, but rather software designed to look like it’s thinking like a human.

As they go on to explain, AI is programmed precisely to mimic four characteristics of human cognition: learning, reasoning, self-correction, and creativity. AI follows rules “referred to as algorithms” that guide the software to follow “precise, step-by-step instructions for accomplishing specific tasks,” the three authors explain.

Thus, AI “learns” by following algorithmic instructions. It can “reason” in the sense that an AI program focuses on those algorithms most suited to a task it’s been asked to perform. AI can then “self-correct” to the extent that it’s been programmed to “continuously refine algorithms” to come up with results that best resemble millions of real-world examples. And via this self-correcting function, AI can create original work, or at least what looks like original work. This “creativity” is still, however, based on data AI programs are trained to rely on from the internet (which is why most journalists still consider this “original” work to be plagiarism when it’s used to write articles).

Artificial extension service

The European Parliamentary Research Service dives deep into the topic of artificial intelligence and agriculture in a 2023 report. The EU report concludes that AI can be used in virtually every aspect of modern, mechanized farming, but only if every corner of these farms is connected to the internet.

“Bringing the benefits of AI and digital agriculture to all farmers requires accessibility to networks and affordable broadband internet access,” the report says, “not only in residences but also in the fields.” This poses a major problem for smallholder farms in developing countries.

If this obstacle can be overcome, then some researchers believe AI programs could be retooled as potentially valuable automated extension services for smallholder farmers.

AI’s promise and potential in farming

Advanced AI tools could be designed to “learn” from data they acquire from farms.

A program might be used to soak up data from a farmer’s field, make note of crops’ conditions, and cross-reference this data against weather and rainfall conditions, temperature readings, and other variables to output recommended steps the farmer could take to ensure a healthy, wealthy harvest. That’s the argument Jayasingh, Anand, and Das make in their analysis. But they acknowledge that, for the moment, such technology is more likely to be used by government extension workers who may then deliver this AI-enhanced agricultural advice to farmers in the fields.

An even bigger hurdle, they say, could be farmer acceptance.

Even if innovation can find a way to close the AI gap between professional consultants and smallholders, farmers who are used to relying on traditional knowledge may not be receptive to extension workers’ help, especially when it’s derived from machines.

Cost is another consideration, one of the reasons why, for now, speculation of artificial intelligence’s place in smallholder farming is just that—speculation. Search for concrete examples of how chatbot-like technology is helping smallholders grow food today and you’ll come up empty.

“Challenges include the high cost of AI installation, resistance from farmers, the need for digital education, potential technical glitches, and the irreplaceable human touch in extension services,” the three Indian researchers wrote. Nevertheless, “AI’s potential to reshape agriculture is undeniable, making it a crucial tool for the future of farming,” including smallholder farming.

 — Grow Further

Photo credit: Kweke Photography 

Subscribe to Our Newsletter