An intelligent approach: Integrating ChatGPT for experiment planning in biochar immobilization of soil cadmium

Abstract

Immobilization of cadmium (Cd) in soil is a complex systematic process in which biochar materials and experimental conditions need to be selected from an infinite space of candidates. Exploration of the entire space is not feasible due to the large number of repeated experiments required. In this study, we propose a strategy for deeply integrating AI techniques to determine the biochar properties and experimental conditions for achieving maximum immobilisation rates under specific soil conditions. We developed six interpretable tree models, among which the LSBoost model performed best. In addition, we developed a Graphical User Interface (GUI) through which soil properties can be manually entered and Bayesian algorithms applied. This allowed us to inversely extrapolate the experimental conditions that achieved maximum carbon sequestration rates under these soil conditions. Then, we use these conditions to link with OpenAI and utilize ChatGPT to obtain detailed experiment plans, which are displayed on the GUI interface for researchers' reference.