Robert Pearl, a professor at Stanford medical school and prior CEO of Kaiser Permanente, a network covering 12 million patients in the US, has proclaimed the importance of ChatGPT, a generative AI system, in the physician’s practice. He believes that it’s impact is analogous to the significance of the stethoscope in the past. Pearl no longer practices medicine, but he knows there are physicians who use ChatGPT to summarize patient medical data, compose medical letters, and ask for advice on diagnoses. He thinks that it will still be able to be used to benefit humans in many ways.
Software companies have taken it upon themselves to devise language models to complement Google search and improve the industry. Doctors have speculated the potential use of this technology could be in the digital health records, providing patients with readable summaries of detailed medical notes. Although they expect the technology to improve medical care, they also fear that it might misguide doctors, providing incorrect diagnoses that could put their patients in harm’s way. Companies such as Microsoft Research and DeepMind have created a standard for the competition to make language models more accurate through medical school exams.
Heather Mattie, a lecturer in public health experienced ChatGPT first-hand when she used it to request information on how modeling social connections is utilized to study HIV. Unfortunately, the model went beyond her knowledge and she couldn’t distinguish accurate information from false. She contemplates the capability of the dataset to discern between two opposing conclusions, and how the user recognizes whether the response is suitable or hazardous. Even though she is more optimist now than when she initially used it, Mattie suggests that it shouldn’t be used in a clinical setting as it doesn’t make clear when it is providing specifics from older records.
OpenAI’s ChatGPT has been a great asset to medical practice, providing medical professionals with sources to expedite patient care and get more accurate records. With AI tools growing more and more advanced each day, programmers, doctors, and lawyers alike have the potential to rely more and more on algorithms to sort through and provide information in ways we couldn’t even have imagined just a few years ago. But caution must be maintained, as there is always the potential of bias and inaccuracy in the tool.
