AI medicine comes in two versions and most people are using the wrong one when they are sick at midnight.

Specialist AI in oncology, antibiotic resistance research, and pandemic prevention is genuinely promising. Greg Fish writes about all of it. The problem is that general chatbot AI, the kind you open on your phone when you feel terrible and your doctor’s office is closed, operates completely differently. It goes through a checklist, has no curiosity, does not know when it does not have enough information, and gets the diagnosis wrong approximately 80 percent of the time on the first attempt.

To demonstrate the credulity problem, scientists created a fake skin condition called bicsonemia, attributed the research to the Professor Sideshow Bob Foundation and the Fellowship of the Ring, published it as blog posts, and watched chatbots pick it up uncritically and diagnose patients with it. No peer-reviewed research. No verification step. Just pattern matching from whatever is available.

Greg Fish says the answer is not to avoid AI in medicine. It is to understand that a human doctor who is curious, makes value judgments, and asks follow-up questions is still required to deploy the tools correctly.

Topics: AI medicine, chatbot diagnosis, specialist AI, medical AI risks, cyberpunk survival guide

GUEST: Greg Fish | cyberpunksurvivalguide.com

Originally aired on2026-05-06