Generative AI for Epidemiologists

Generative AI tools, such as Large Language Models (LLMs), represent a subset of artificial intelligence technologies capable of generating new content, including text, images, audio, and video, that mimics human-generated content. Unlike earlier AI approaches, such as machine learning, which are designed to recognize or classify data, generative AI can create novel data outputs by leveraging the patterns, styles, or information it has learned during its training process. This ability to recognize and generate patterns shares many similarities with the goals of Epidemiology, which focuses on identifying patterns of health and disease in populations.

Although these generative AI models are relatively new, their adoption in the research environment, including in Epidemiology, is rapidly increasing. The introduction of LLMs has the potential to revolutionize scientific research by providing unprecedented speed, innovation, and efficiency. This is achieved by enabling novel and compelling ways to explore data. However, the complexity of the neural networks behind these models' decision-making processes and lack of transparency in data training sources, can make it challenging to understand how they arrive at specific outcomes and biases. The opacity of these models, coupled with the necessity of training them on potentially biased data sets, underscores the need for responsible use. Epidemiologists, therefore, face the challenge of ensuring that research findings generated by AI are real, ethical, and reliable.

This course is designed to introduce LLMs, highlighting their potential to enhance epidemiological research. It aims to explore innovative ways LLMs can be utilized, understand the myriad ethical considerations involved, investigate the potential public health concerns raised by AI, learn how to conduct basic analyses, and examine firsthand applications of LLMs within the field of epidemiology. The class will include informational lectures with in-class applied discussions and laboratory learning exercises using LLMs.