According to Andrew Beam of Harvard TH Chan School of Public Health, artificial intelligence (AI) has the power to change medicine and public health for the better. But using this powerful tool also requires caution, he said.
An assistant professor of epidemiology, Beam plays a key role in two collaborative efforts by the NEJM Group (publisher of the New England Journal of Medicine) investigating the use of AI in medicine. He, along with his Arjun (Raj) Manrai at Harvard Medical School, are co-deputy editors of his NEJM AI, a journal that will publish its first articles this fall. Isaac (Zak) Kohane, founder of the Department of Biomedical Informatics at the Bravatnik Institute at Harvard Medical School, will serve as editor-in-chief of NEJM AI.
Additionally, Beam and Manrai co-host a podcast called NEJM AI Grand Rounds. In this podcast, our guest expert explores deep issues at the intersection of his AI, machine learning, and medicine.
Beam said the launch of the journal and podcast came at a critical time.
“AI used to be a hypothesis of the future. But in the last two years, things have changed,” he said. “AI is here. Everyone knows about ChatGPT. But healthcare is different because patient lives are at stake. Medical AI is more risky than other types of AI That’s what really spurred me, Zack, and Raj to come to this journal — to really kick the tires and get serious about where to figure out when medical AI can be used safely. .”
The first three episodes of the podcast focus on three different areas of medicine. The first is a feature where Ewan Ashley of Stanford University School of Medicine talks about his AI, genomics, and cardiology. Second, his Pranav Rajpurkar from Harvard Medical School explains his AI and Radiology. Third, Verily’s Lily Peng talks about AI in ophthalmology. In his latest episode, Microsoft Research head Peter Lee speculates on how technologies like ChatGPT will transform healthcare.
“[Ashley and his colleagues are] We ask questions like, ‘Can we have an AI read an echocardiogram like a human cardiologist does?’,” says Beam. “They also found that AI can extract signals. [from echocardiograms] Things people can’t do, such as biomarkers that indicate anemia. “
Peng spoke about the potential public health benefits of the technology. Her team conducted trials in Thailand, where there are few trained ophthalmologists, and found that AI works well in screening for diabetic retinopathy in that population, helping people get diagnosed and treated more quickly. He said he found it possible.
Aside from its promise, Beam also acknowledges that AI poses serious challenges when used in medicine.
“Systems like ChatGPT have structural biases built in. This is because the data is trained. The data is literally the entire internet and always reflects our best selves.” Not necessarily,” said Beam. “So we need AI to operationalize these biases at scale and not hurt people who are already on the fringes of the health system.”
— Karen Felscher
Read full text