Artificial Intelligence and Machine Learning May Resolve Health Care Information Overload
Mark G Siegel,Michael J Rossi,James H Lubowitz
DOI: https://doi.org/10.1016/j.arthro.2024.01.007
Abstract:Biomedical information doubles almost every 2 months, and this very rate is expected to double by 2025. The result is information overload for clinicians and researchers. Today, artificial intelligence (AI) and machine learning (ML) research contribute to the deluge of information. In addition, AI large language models, although capable of automating scientific writing, are flawed. They hallucinate (make things up), are trained primarily on non-peer-reviewed content, raise ethical and legal issues, and lack human empathy. Still, when it comes to AI including ML, we are optimistic. The technology is improving rapidly. In the future, AI will help us manage unwieldy information by processing data, determining diagnoses, recommending treatments, and predicting outcomes. In research, AI and ML similarly promise efficient data analysis and literature review and will create new content in response to our instructions. Human touch will be required, and we will disclose use of AI proactively, including rationale for its use, our data input, our level of confidence in the output, and the patients or populations to whom the output may be applied. In addition, we will ensure data quality is high and bias is minimized. Most of all, we will provide essential reasoning, clinical and research guidance, and diligent oversight. Humans will remain accountable.