Why general-purpose AI falls short in agriculture

TechRadar
Read full story· 6 min
Researchers tested generative AI chatbots like GPT-3.5 and GPT-4 for agricultural advice and found they often provide incorrect or ambiguous guidance, risking crop damage and food security. Farmers increasingly rely on AI due to shortages of professional advisors and rising costs, but current models fall short in delivering accurate, localized farming recommendations.

What happened

In July, a farmer in Chuzhou, China, used a generative AI chatbot to get advice on controlling weeds and pests. The chatbot suggested a mix of chemicals, which the farmer applied to his 25-acre sesame field, resulting in the loss of all his crops and a financial loss of about 150,000 yuan. Researchers from several countries later tested AI chatbots like GPT-3.5 and GPT-4 for agricultural guidance and found that these models often gave ambiguous or incorrect advice. Many farmers are turning to AI for help due to a shortage of professional advisors and rising costs, but current general-purpose chatbots struggle to provide accurate, localized recommendations.

Key details

  • The farmer lost 25 acres of sesame crops, worth about 150,000 yuan (18,000 euros), after following AI advice.
  • Researchers from the U.S., U.K., Kenya, Nigeria, and Colombia tested GPT-3.5 and GPT-4 for agricultural advice.
  • Chatbots gave ambiguous or incorrect recommendations for pest and weed control, including wrong herbicide timing for cassava in Nigeria.
  • About 17% of farmers worldwide use generative AI for daily decisions, according to a recent survey.
  • Shortages of trained agricultural advisors and high costs are pushing more farmers to rely on AI tools.

Why it matters

Farmers who depend on general-purpose AI chatbots for crop management face a risk of receiving harmful or inaccurate advice, which can lead to significant financial losses and threaten food security. The lack of access to professional agricultural advisors, especially in remote regions, means more farmers may turn to these tools despite their limitations. This situation creates an urgent need for specialized AI systems that use real-time, field-specific data to provide safe and effective guidance.


More in Food & Agriculture