Former CEO of Kaiser Permanente, a US healthcare organization with more than 12 million patients, is now a professor at Stanford Medical School.
His name is Robert Pearl. If he were still in charge, he would demand that all of the 24,000 doctors there immediately begin utilizing ChatGPT in their practices.
According to Pearl, “I believe it will be more significant to doctors than the stethoscope was in the past.” No doctor who provides high-quality care will do it without using ChatGPT or other generative AI tools, according to the statement.
Even though Pearl no longer practices medicine, he claims to know doctors who use ChatGPT to write letters, describe patient care, and even – when stumped — ask for suggestions on how to diagnose patients. He believes that doctors will find tens of thousands of beneficial uses for the bot to advance human health.
Language models are beginning to demonstrate the capacity to take on duties historically designated for white-collar specialists like programmers, attorneys, and doctors as technologies like OpenAI’s ChatGPT challenges Google search’s dominance and sparks discussions of industry revolution. That has generated discussions among physicians about how technology can improve their ability to care for patients. However, there is also concern that language models could mislead clinicians or give wrong answers that would result in an improper diagnosis or treatment plan. Medical practitioners believe language models can uncover information in digital health records or give patients summaries of extensive, technical notes.
Medical school exams have become a standard for AI technology developers as they compete to create systems that are more powerful. A report by OpenAI, Massachusetts General Hospital, and AnsibleHealth stated that ChatGPT can equal or exceed the 60 percent passing score of the US Medical Licensing Exam. Last year, Microsoft Research unveiled BioGPT, a language model that scored highly on a variety of medical activities.
A few weeks later, researchers from Google and DeepMind unveiled Med-PaLM, a system that scored 67 percent correctly on the same exam. Despite this, they noted that while encouraging, their findings “remain inferior to clinicians.” In order to find trends in electronic health data, Microsoft and Epic Systems, one of the major providers of healthcare software, have announced intentions to employ OpenAI’s GPT-4, which powers ChatGPT.
The first time Heather Mattie tried ChatGPT, she was impressed. Heather Mattie is a Harvard University lecturer in public health who researches the effects of AI on healthcare. She requested an overview of how modeling social networks has been applied to explore HIV, a subject she studies. The model eventually moved into areas she knew nothing about, and she was unable to tell if it was true. She found herself questioning who decides whether a response is appropriate or damaging, as well as how ChatGPT reconciles two radically different or competing results from medical articles.
Since that early event, Mattie says she has become less gloomy. She claims that if the user is aware that the bot might not always be accurate and sometimes produce biased results, it can be a valuable tool for tasks like summarizing content. She is concerned about how ChatGPT handles cardiovascular disease and intensive care injury score diagnostic tools, both of which have a history of racial and gender bias. However, she is still wary about using ChatGPT in a clinical context because it occasionally makes up facts and doesn’t make it clear when the data it is using is current.
There is no way to identify where in the history of medicine ChatGPT gets its information from when reporting a typical treatment because medical knowledge and practices vary and advance throughout time. Is that knowledge current or out of date?
Users should also be cautious since ChatGPT-style bots may offer made-up, or “hallucinated,” material in an apparent fluent manner. If a user doesn’t fact-check an algorithm’s responses, this could result in major mistakes. AI-generated text can also subtly affect people. According to a January study that used ChatGPT to answer ethical questions, which was not peer reviewed, the chatbot is an erratic moral advisor and can have an impact on people’s decisions even when they are aware that the advice is coming from AI software.
There is much more to being a doctor than simply reciting encyclopedic medical knowledge. While many medical professionals are excited about using ChatGPT for low-risk tasks like text summarization, some bioethicists are concerned that doctors will consult the bot for guidance when they must make a difficult ethical choice, such as whether surgery is the best course of action for a patient with a low chance of survival or recovery.
Jamie Webb, a bioethicist at the Center for Technomoral Futures at the University of Edinburgh, asserts that “you can’t outsource or automate that kind of process to a generative AI model.”
In response to earlier studies that proposed the notion, Webb and a group of moral psychologists investigated what it would take to create an AI-powered “moral adviser” for use in medicine last year. According to Webb and his coauthors, it would be difficult for such systems to consistently balance various ethical values, and doctors and other staff members would experience “moral de-skilling” if they become unduly dependent on a bot rather than using their own judgment to make difficult judgments.
Webb notes that previous claims that language-processing AI would change medicine have left doctors unimpressed. The Watson division at IBM went to oncology and claimed that using AI to fight cancer was beneficial after winning Jeopardy! in 2010 and 2011. However, the Memorial Sloan Kettering in a box solution wasn’t as effective in clinical settings as the marketing hoopla would have you believe, and IBM ended the project in 2020.
When hype is hollow, there might be long-term repercussions. Primary care physician Trishan Panch reported seeing a colleague tweet the findings of asking ChatGPT to diagnose an ailment shortly after the chatbot’s launch during a discussion panel on the possibilities of AI in medicine at Harvard in February.
Excited medical professionals pledged to apply the technology right away, Panch recalled, but by the 20th comment, another expert chimed in and declared that every reference produced by the model was false. “It only takes one or two things like that to erode trust in the whole thing,” asserted Panch, cofounder of the health care software startup Wellframe.
Robert Pearl, a former employee of Kaiser Permanente, is still quite optimistic about language models like ChatGPT despite AI’s occasionally apparent errors. He predicts that language models in healthcare will resemble the iPhone more in the coming years, being loaded with capabilities and power that may support medical professionals and assist patients in managing chronic diseases. He also has some doubts that language models like ChatGPT can lessen the more than 250,000 deaths brought on by medical mistakes each year in the US.
Some things are, according to Pearl, off-limits to AI. He asserts that using a bot to assist with end-of-life discussions with families, assist people in coping with grief and loss, or discuss procedures with a high risk of complications is inappropriate because each patient’s requirements are so different and require individual discussion.
“Those are human-to-human conversations,” adds Pearl, noting that the current state of technology only represents a small portion of the potential. “If I’m incorrect, it will be because I underestimate how quickly technology is developing. However, it’s moving faster than I even realized every time I look.
For the time being, he compares ChatGPT to a medical student who can help out and care for patients, but everything they do needs to be approved by a practicing doctor.
Download The Radiant App To Start Watching!
Web: Watch Now
LGTV™: Download
ROKU™: Download
XBox™: Download
Samsung TV™: Download
Amazon Fire TV™: Download
Android TV™: Download