This work's primary contributions are:

> Conducting a preliminary study on ChatGPT's potential in agricultural text classification and proposing ChatAgri, a ChatGPT-based solution.
> Evaluating ChatAgri on multiple multi-lingual datasets and demonstrating its competitive performance compared to existing PLM-based fine-tuning approaches, including producing a reasoning chain that simulates the human mind.
> Demonstrating ChatAgri's potential in agricultural text classification through zero-shot setting experiments, which require less supervised data, annotations, and domain expert knowledge.
> Showing ChatAgri's excellent domain transferability in multi-lingual experiments, making it adaptable to different agricultural application scenarios and accelerating the future of Artificial General Intelligence (AGI).
> Subverting the mainstream complex and power-intensive PLM-based methods, ChatAgri only relies on network interface and minimum hardware requirements, making it a promising low-cost artificial intelligence technique for future smart agricultural applications.
> Releasing the codes of ChatAgri on Github and deploying a ChatGPT-based agricultural text classification system to an online website to encourage further research on smart agricultural applications using ChatGPTs.