Key points from article :
AI tools have been speculated to play a role in healthcare, with some outperforming physicians in specific tasks.
Concerns over AI replacing specialists and exposing patients to risks have diminished but reemerged with language learning models (LLMs) like OpenAI's GPT.
GPT is a generative pretrained transformer, producing human-like text and enabling interaction with AI.
GPT's potential applications in healthcare include research, education, and clinical care.
In clinical practice, GPT can assist with repetitive tasks, provide decision support, and be incorporated into electronic medical records.
Current GPT versions are not HIPAA compliant, posing a risk to patient privacy.
Legal and ethical questions surrounding GPT use depend on how it affects social relationships between users.
AI within the patient-physician relationship that augments rather than replaces clinician judgment poses fewer risks.
Patient-facing AI in healthcare delivery that substitutes for clinician judgment can undermine trust and competence in the patient-physician relationship.
Direct-to-consumer health advice-giving is the least regulated use case, with few legal guardrails.
Clinicians should educate patients about the risks of using LLMs like GPT outside the patient-physician relationship.
Until proven accurate and reliable, GPT should not replace clinician judgment.
Clinicians could advocate for Federal Trade Commission (FTC) regulation to protect patients from false or misleading AI-generated medical advice.
Viewpoint by Claudia Haupt and Mason Marks published in JAMA
-thumbnail-AC.jpg)
