Dr. Adam Rodman, a respected physician and researcher at Beth Israel Medical Center in Boston, was astounded by the findings of an experiment he conducted comparing doctors’ diagnostic skills to those of generative artificial intelligence (genAI) tools. The results revealed that even with technological assistance, physicians were significantly less accurate in diagnosing patients compared to genAI operating independently.
“Upon repeating the experiment, I was truly taken aback,” shared Rodman, who also serves as the Director of AI Programs at Beth Israel. “The AI performed nearly 20% better on its own, achieving an impressive accuracy rate of around 90%.”
“We had assumed that human doctors would outperform AI systems, so it was quite surprising to discover that not only did the technology fail to enhance physicians’ performance but also surpassed them significantly in diagnostic accuracy,” he added.
The genAI model utilized in the study, GPT-4 Turbo from OpenAI – the same technology behind Microsoft’s Copilot chatbot assistant - not only outshone human doctors but also surpassed all other healthcare-focused AI systems developed over the past five decades without any specialized medical training.
Conducted by Rodman and Dr. Jonathan Chen, an esteemed Assistant Professor of Medicine at Stanford University in 2023; it is common for research findings to be published after a considerable period following completion.
Rodman hinted at even more groundbreaking results from another upcoming study set for publication in Nature within two months. ”We are actively developing next-generation models and AI tools aimed at enhancing physicians’ capabilities. Stay tuned for more exciting studies on this front,” he teased.
2024-12-13 23:15:02
Link from www.computerworld.com