May 21, 2024 - Health

Newest version of ChatGPT could accelerate chatbot care

Illustration of a caduceus with arrow cursors for wings

Illustration: Sarah Grillo/Axios

The latest iteration of OpenAI's ChatGPT, with its sharper real-time voice interactions, may mark a turning point in how people rely on chatbots for health care.

Why it matters: Doctors say the new model's humanlike qualities could make patients more comfortable querying the generative AI chatbot for health care advice.

  • On one hand, doctors say that could make patients more engaged in managing their health. But they also worry that patients will be readier to push aside their advice as these chatbots — which are still prone to offering misinformation — get better at approximating real-life conversations.

Catch up quick: OpenAI last week said it will soon roll out GPT-4o, an updated version of its flagship model that can read facial expressions and understand voice or text commands, even when interrupted.

  • This makes for a faster, more confident version of the chatbot. But that speed also exacerbates a key weakness: It is often also confidently wrong.

Between the lines: It's not hard to imagine how this kind of interactive tool could be used in treating patients, said Robert Pearl, a Stanford University professor and author of the book "ChatGPT, MD."

  • In a demo of its tool, OpenAI showed an employee getting a real-time tutorial on taking deep breaths to calm his nerves, while another showed how the tool could detect his emotions by looking at him.
  • "This is making it possible for you to have conversations identical to what you'd have with your closest friends, to what you'd have with your physician," Pearl told Axios.
  • The tech may eventually evolve into a life-saving tool, offering emergency responders clearer analysis of a patient's condition, providing step-by-step assistance to surgeons in the operating room or offering empathetic support for mental health issues, wrote Harvey Castro, an emergency medicine physician who's spoken extensively on AI in health care.

Yes, but: Surveys show that patients are wary of doctors relying on AI to treat them, and doctors are less than thrilled about the possibility of being supplanted.

  • But it's less clear how patients view the technology for managing their own health.
  • Providers who spoke with Axios raised concerns that patients may put too much confidence in the chatbots' capabilities. They also said patients wary of going to a doctor, or concerned about affording one, may see these tools as an acceptable alternative.
  • However, the tech won't necessarily have a patient's health history or offer nuanced care that a human provider would, providers say.
  • "People already go to Dr. Google," said Sarah Oreck, a psychiatrist and CEO of Mavida Health. "I fear that will become even more robust."

Still, not all doctors worry about being replaced. Pearl, who previously led Kaiser Permanente's physician group, said he envisions a complimentary role for the tech.

  • When doctors recommend patients make lifestyle changes such as diet and exercise improvements, he said they might also suggest consulting a chatbot for help developing the best plan for them.
  • "It will provide you with all the pieces that you need, the shopping list, and the recipes, it will give you the exercise and it will serve as a health coach," he said.
Go deeper