Nov 29, 2023 - Health

A year after launching, ChatGPT is already changing medicine

an illustration of a circuitboard with a red-cross-shaped section illuminated

Illustration: Tiffany Herring/Axios

It's passed medical licensing exams. It's advanced how researchers develop new medicines and cut down on doctors' hefty paperwork. And it's nudged health care closer to a world where AI can offer diagnoses.

Why it matters: One year after OpenAI's ChatGPT exploded onto the scene, the generative AI model is already upending health care — an industry not exactly known for its speedy adoption of tech — while accelerating questions about AI's promises and limitations.

The big picture: While AI and algorithms have been used in health care for decades, ChatGPT and other generative AI models that quickly followed have supercharged their use across research and the delivery of care.

  • "All these things that are happening now [are] because the race to get to AI faster, quicker and to create the market share began with ChatGPT," said Shafiq Rab, system CIO and chief digital officer at Tufts Medicine.
  • "We're in the 'wow' period," Stanford's School of Medicine dean Lloyd Minor told Axios at a health tech event earlier this fall.
  • He and other experts predicted that the unprecedented hype around how AI may change health care will begin to quiet down over the next few months as the industry races to get a better handle on what the technology can and can't do.

Catch up quick: Early in 2023, experts were holding "conversations" with ChatGPT on subjects like stem cell research or the future of surgery.

  • They quickly graduated to testing how well ChatGPT performed on the U.S. Medical Licensing Exam. (It passed.)
  • Soon, clinicians started wondering if it could lighten their workloads, with studies examining how well ChatGPT wrote discharge summaries or radiology reports. In some cases, it was even found to be more empathetic in answering patient questions.
  • With the rollout of other large language models from competitors such as Google and Amazon, tech companies started looking at how to help doctors more efficiently sift through the latest research or listen in on patient visits to give doctors a hand with documentation.
  • The pharma industry is already using generative AI models to make drug discovery more efficient. Tech giants have been finding ways to use the algorithms to better target cancer.
  • "What we did in the last 100 years, we'll achieve in the next 10 years, or even five years," Rab predicted.

Yes, but: There's also been quite a bit of hand-wringing about AI chatbots' potential to churn out harmful misinformation, amplify bias and undermine medical education.

  • The medical community has emphatically dismissed the idea that AI could replace doctors' role in diagnosing patients — but research is already raising questions about its potential.
  • While the evolution of generative AI has been hailed for its potential ability to develop biodefense tools, there have also been warnings about its ability to create new biological threats.

Our thought bubble: At a recent health tech conference, it seemed every other company and health system there touted their use of AI.

  • It was a clear sign health care was in the throes of the generative AI hype cycle, something several folks I interviewed there also observed.
  • "It's powerful. But I think you know, the hype right now is, is probably a little more than reality," Headspace CEO Russ Glass said about the number of companies boasting about AI since ChatGPT's launch. His company uses AI to assist behavioral health clinicians with their notes and responses to patients.
  • "We don't see a time in the near future that it's going to supplant human intervention. But it is important," he told Axios.
Go deeper