AI’s Limitations in Clinical Decision-Making (Medscape)

A recent study in npj Digital Medicine explored the potential and limitations of GPT-4V, a multimodal AI model, in healthcare diagnostics. While the AI tool outperformed doctors in accuracy (81.6% vs. 77.8%) on a set of diagnostic questions, it struggled with image interpretation and providing reliable justifications for its diagnoses. The findings emphasize that AI can support but not replace human expertise, highlighting the need for cautious integration of AI into clinical practice and further research to understand its limitations and potential in real-world scenarios.