China launches AI model for plant protection
BEIJING (ANTARA) - Science and Technology Daily reported on Tuesday, 26 May, that China has launched its first open-source large language model (LLM), Green Shield, to enhance plant protection.
Developed by Nanjing Agricultural University (NAU) in collaboration with the National Key Laboratory of Agricultural Biosafety and over 30 industrial institutions, the model is designed to provide science-based agricultural guidance and ensure proper pesticide use.
“China frequently faces plant pest outbreaks and pesticide resistance issues,” said Dong Shameng, project leader and NAU’s Vice Dean of the Plant Protection Faculty.
Dong, quoted by the newspaper during the launch event on Monday, 25 May, added: “Farmers urgently need professional grassroots guidance. However, general-purpose LLMs often provide inaccurate answers to plant protection queries and, more critically, offer non-standardised, even risky, pesticide usage advice.”
To address this, the team built a specialised corpus of over 2.5 billion tokens sourced from academic papers, patents, national standards, and field reports.
According to the report, the corpus covers major crops including rice, wheat, soybeans, vegetables, and fruit trees, integrating information on pest monitoring, eco-friendly control measures, and pesticide registration.
Wang Dongbo, a professor at NAU’s Faculty of Information Management, said the model can precisely identify plant types, growth stages, and disease symptoms, then generate integrated growth control strategies.
“With targeted training, this model performs well and can identify pests with high precision,” Wang stated.
Wang explained that before offering any recommendations, the model automatically cross-checks against the national pesticide registration database to verify chemicals against banned lists, approved crops, and dosage limits. Any non-compliant advice is blocked and corrected automatically, preventing pesticide misuse from the outset.
NAU Vice President Wang Yuanchao said the university would continue field testing and model iterations to create a “user-friendly, practical, and effective” smart tool for farmers, empowering modern agriculture with digital technology across the production chain.