Abstract
Artificial intelligence (AI) has broadly reshaped health and medicine, benefiting clinicians, patients, and health systems. However, technical, regulatory, and ethical challenges exist in the application of medical AI, ranging from data scarcity to fairness. We provide our perspective on how to address the major challenges facing widespread clinical adoption from both technical (e.g., building high-quality datasets, using larger and more diverse datasets for training, creating problem formulations that go beyond supervised learning, and combining human skills with AI tools) and ethical (e.g., using highly secure data platforms and strengthening governmental legislation) perspectives.