In this episode of Healthcare Corner, Dr. Joseph Arrington, who holds a Ph.D. in population health sciences from the University of Utah, explains why people use LLMs like ChatGPT, Gemini, and Claude for health decisions and the serious risks involved. He says that these models predict text patterns rather than reasoning like clinicians, which can lead to confident but incorrect guidance on health matters.

Comparing this to the Dunning–Kruger effect, Dr. Arrington shares how even experts can be misled. He highlights real-world dangers, such as the lawsuit from Pastor Scott Winters, and notes that LLMs lack HIPAA protections, cannot perform physical exams, and may reinforce training biases. While AI should never replace professional medical judgment, it can safely be used to explain medical terms, organize symptoms, and prepare questions for appointments.

Support TheLine.Info:

https://buymeacoffee.com/theline.info
https://ko-fi.com/thelineinfo

Be sure to like and subscribe!

Follow TheLine.Info:

Instagram: @theline.info
Facebook: Facebook.com/theline.info
BlueSky: @thelineinfo.bsky.social

Check out the website:

https://thelineinfo.wixsite.com/main

Recommended video: "Vibe Physics" by Angela Collier
vibe physics

CHAPTERS

00:00 Introduction/Disclaimer
00:18 Welcome back
00:35 What are LLMs (Language Learning Models)?
00:44 How people have used LLMs in the past
01:18 The Allure of AI
02:42 Predicting words, not health
04:04 The false confidence trap
04:24 Dunnig-Kruger with Reddit examples
07:41 The limits of A.I.
08:54 WebMD
09:31 The sci-fi novel example
10:46 What a headache means
11:53 Prompts
12:10 Giving a random person a TI-84 plus CE graphing calculator
15:19 Underdetermination
17:09 Why people go to medical school
17:54 The Scott Winters story
19:26 WebMD on steroids
20:34 HIPAA compliance
21:15 Bias and Privacy
27:28 How to properly use A.I. for health information
30:05 Final thoughts
30:31 End credits