By Alexis Kayser
Healthcare Editor
0
Share
Newsweek is a Trust Project member
See more of our trusted coverage when you search.Prefer Newsweek on Googleto see more of our trusted coverage when you search.
Microsoft AI unveiled a dozen new features as part of its Copilot Fall Release on Thursday—including some that aim to provide more accurate, useful responses to health queries.
When users come to Copilot for health-related help, they’re typically searching for health care providers or researching medical conditions that they or a loved one might have, Dr. Dominic King, vice president of health at Microsoft AI, told Newsweek. The chatbot has been updated to address both more comprehensively, with more trustworthy information.
Read More on Health Care
“We're seeing about 50 million health sessions a day on Microsoft's consumer services, Copilot and Bing,” King said, “and we feel a heavy responsibility to do a very good job there.”
The popular large language model (LLM) will now offer credible sourcing for health-related questions through a partnership with a world-renowned source of medical expertise. Harvard Health Publishing (the health information arm of Harvard Medical School) collaborated with Microsoft AI to provide accurate information with references to Harvard Health articles, so that users can place more trust in the chatbot’s recommendations and trace them back to a validated source.
Microsoft AI has also grounded Copilot in a specific set of “credible” sources as defined by the National Academy of Medicine. When users ask a health question, the chatbot will only draw from vetted organizations like Mayo Clinic, Cleveland Clinic, Kaiser Permanente and other trusted national health systems.
This is an improvement from the previous version of Copilot, which cast a much wider net to generate answers, according to King.
“What we see when we do a standard internet search, whether that's on Bing or other surfaces, or engage with a conversational AI, that information tends to be sourced from a really wide variety of sources, almost the entire internet,” King said. “Obviously amongst that, you'll find very reputable, credible sources, and also less credible sources that may include blog posts or Reddit posts and various other things.”
“When you see a generative answer [on Copilot], that will now be sourced from trusted places and very carefully referenced,” King continued. “You see the answer and you also see where it's come from.”
In this demo video from Microsoft AI, watch how Copilot answers a user's health questions with the new credible sources and care navigation features.
The fall update also equips Copilot to assist users with health care navigation. For example, a user could ask, “Find cardiologists in my area,” and receive a response including local physicians’ locations, clinic hours, accepted insurance plans, spoken languages and specialties.
Microsoft AI sourced this information from health care providers to bring more accurate data to users, King said. He hopes that this will reduce “friction” in patients’ search for physicians, as much of the existing information online is out of date.
“When you search competing products, you may see operating hours [but] whether they're correct or not [is] questionable,” he said. “I think what you'll see from what we're providing now is a much higher level of accuracy in that information, and a much broader range of information being presented.”
There is one more new feature that may prove helpful in health care scenarios, King said. Copilot now offers “Groups,” allowing users to invite up to 32 people to collaborate in real-time. It can “summarize threads, propose options, tally votes and split tasks” to make communications more efficient, according to a Thursday blog post from Microsoft AI.
“There's a lot of discussion at the moment about, how do you manage care for children, adults and elderly parents at the same time?” King said. “So group chatting features like [Copilot’s] could become pretty compelling for health, too.”
View 0 Comments
Submit Correction
View Editorial & AI Guidelines