Using ChatGPT to Predict the Future of Diabetes Technology

The artificial intelligence (AI)-based chatbot called Chat Generative Pre-trained Transformer (ChatGPT) has gained much popularity across the internet. Developed by OpenAI, ChatGPT is an open source language model trained using reinforcement learning from human feedback. ¹ The goal of ChatGPT is to generate factually accurate and linguistically coherent responses to user questions. ¹ Although the source of the dataset used to train ChatGPT is not disclosed by OpenAI, it has been speculated that the model is trained using text databases on the internet including books, articles, Wikipedia, and the WebText dataset.^{2 -5}

A literature review conducted on PubMed on February 15, 2023, yielded 40 relevant articles on the use of ChatGPT in medicine, including academic publishing, medical education, data interpretation, patient communication, telehealth, and medical data collection. No article discussed the use or relevance of ChatGPT for diabetes or diabetes technology. Therefore, that day we asked ChatGPT the question: “What is the future of diabetes technology?” The response provided by ChatGPT was as follows:

The future of diabetes technology looks promising, as researchers and engineers are working on developing new and innovative devices and tools to help individuals with diabetes manage their condition more effectively. Here are some of the areas that are showing significant progress:

1. Continuous Glucose Monitoring (CGM): CGMs are devices that continuously measure blood glucose levels and provide real-time feedback to the user. These devices have been around for some time, but the technology is improving rapidly. In the near future, we can expect CGMs to become even more accurate, smaller, and more affordable.

2. Artificial Pancreas Systems: An artificial pancreas system is a closed-loop system that automatically adjusts insulin delivery based on CGM readings. These systems are in development and have shown promising results in clinical trials.

3. Smart Insulin Pens: These are insulin pens that are equipped with sensors to track insulin dosage, timing, and frequency. This can help patients to better manage their insulin regimen and avoid missed or incorrect doses.

4. Non-Invasive Glucose Monitoring: Several companies are working on developing non-invasive glucose monitoring technologies that do not require blood draws. These technologies could be a game-changer for diabetes management, as they would reduce the need for frequent finger sticks and make glucose monitoring more convenient.

Overall, the future of diabetes technology looks very promising, and we can expect to see many new and innovative devices and tools in the coming years that will help individuals with diabetes to better manage their condition and improve their quality of life.

The optimal role of ChatGPT and other chatbot interfaces to answer scientific and clinical questions has not been determined. Although ChatGPT appears to be a promising tool for the development and improvement of diabetes technology, major limitations of the model are the lack of transparency of the data source and the quality of data used to train the model. Rigorous evaluation, data diversity, and quality verification by human experts of the training dataset will be required for ChatGPT to become integrated into research in diabetes.