Applying Prompt Engineering of Generative Artificial Intelligence to the Diagnosis of Diabetes

Abstract

This paper presents a system that relies on artificial intelligence models with enough sophistication built by rigorously adhering to the recommendations of generative artificial intelligence tools such as large language models when prompted to design a system for the automated diagnosis of diabetes. Actual source code for the construction of the artificial intelligence models is generated as part of the suggestions recommended by the generative artificial intelligence tools or large language models. By faithfully incorporating the source code into a module for the automated diagnosis of diabetes based on clinical measurements, adequately sophisticated artificial intelligence models are constructed, trained, tested and validated on publicly accessible diabetes datasets and then deployed in an automated diabetes diagnosis module. Results indicate that the resulting artificial intelligence models exhibit reasonable performance that compares favorably (in view of the fact that the resulting artificial intelligence models are not optimized) with the performance of systems designed from the ground up by artificial intelligence experts.