An AI chatbot with better bedside manner than your doctor? It sounds unlikely, but Google may have pulled it off with a chatbot trained to conduct medical interviews and suggest diagnoses based on medical history.
According to the journal Nature, Google's AI model diagnosed respiratory or cardiovascular conditions as well or better than board-certified primary care physicians, gathered similar information during medical interviews, and even scored higher on empathy.
Google's chatbot is just the latest splash in a tidal wave of AI-based medical software products that claim to do everything from automate routine administrative tasks to predict complications in high-risk patients. Evangelists say that AI will help diagnose complex conditions more easily, make medical care most cost-effective, and guard against potentially deadly human error, according to Forbes. It could even discover new drugs and predict side effects.
But while AI has already gained some popularity as a virtual assistant that cuts back on paperwork, physicians are generally more reluctant to rely on its clinical judgment, according to the New York Times. Their concerns are well-founded — one AI program, designed to predict sepsis, instead inundated doctors with false alarms. And according to a study published in the journal Nature, faulty data set and programming led another AI tool to approve treatments for white patients while denying them to Black patients who were just as sick.
Lawmakers have also sounded alarms about patient privacy, data security, and informed consent. In an October letter to Google CEO Sundar Pichai. Sen. Mark Warner of Virginia expressed his concern that tech companies may be prioritizing market share over patients, and urged Google to be more transparent and take steps to safeguard patient information.