Podcast Summary: "Will AI Take Your Job?"
Derms on Drugs, Scholars in Medicine
Episode Date: January 23, 2026
Main Theme
This episode of "Derms on Drugs" dives into the transformative impact of artificial intelligence (AI) on dermatology practice and healthcare at large. Hosts Dr. Matt Zirwas, Dr. Laura Ferris, and Dr. Tim Patton welcome Dr. Farah Kmanger, founder of DermGPT, to discuss recent research, practical use cases, and professional anxieties around whether AI might replace—or simply augment—the dermatologist's role.
Key Discussion Points & Insights
1. AI vs. Dermatologists: Recent Evidence
- Paper Review: "Comparing Physician and AI (ChatGPT-4) Responses to Common Patient Questions" (Lewandowski, Poland, 2023)
- ChatGPT-4 outperformed dermatologists on patient-rated quality, empathy, and satisfaction regarding answers to questions about hidradenitis suppurativa (HS).
- Notably, 88% of patients preferred ChatGPT's answers, but 70% still said they would rather hear from a human dermatologist.
- Physicians also preferred ChatGPT’s responses.
- Quote:
“ChatGPT-4 kind of crushed the dermatologist. Patients rated the quality, empathy and satisfaction...higher…88% of the time patients said, I like the ChatGPT-4 response better. So...I’ll be out of jobs pretty soon. And I for one, welcome our new chatbot overlords.”
— Tim Patton (03:26) - Discussion highlighted that while AI provides better responses, inherent trust and comfort in human interaction persist… for now.
2. Why Do People Still Prefer Human Doctors?
- The panel likens AI skepticism to early reactions to Uber or self-driving Teslas: unfamiliarity breeds distrust, but attitudes shift as technology matures.
- Dr. Kmanger points out:
- The adoption curve: innovators vs stragglers.
- Human interaction is still prized, though this may shift as AI accuracy and familiarity grow.
- AI is intentionally designed to be empathetic and "super customer servicey."
- Quote:
“It’s intuitive to deal with it. It's language-based and language is how we communicate… these large language models, the reason they're beating out the doctors, they're created to just be super nice, super customer servicey.”
— Farah Kmanger (06:29)
3. Limits and Potentials of AI in Dermatology
-
AI excels at structured communication tasks and longitudinal tracking (e.g., tracking moles via digital dermoscopy).
-
Image-based diagnostic tools are improving rapidly. Devices can now photograph and analyze every mole on the body in minutes.
-
Dr. Ferris:
“Longitudinal follow up of that mole...if that person comes back in six months and it takes another photo...there's no way I'm going to remember what that mole looked like. And even our photography currently is not that good.”
(12:29) -
Over-diagnosis is discussed—AI may help standardize what really needs attention, and allow dermatologists to focus more on complex or procedural cases.
4. AI-Generated Physician Avatars & Patient Acceptance
-
Study from Mayo Clinic: AI-created video avatars of doctors provided post-op patient counseling; patients reported high levels of trust, engagement, and little “eeriness."
-
Acceptability and trust reached 90%.
-
Quote:
“They could now create a digital twin of all four of us… the LLMs will be able to tell the digital twin what to say and patients won't be able to tell it's not us… It is better than us [at empathy].”
— Matt Zirwas (17:16) -
AI avatars and “digital twin” doctors open up 24/7 patient interactions and potentially outperform real doctors on measures of patience and warmth.
-
Dr. Ferris points out more immediate uses like automated counseling and scheduling.
5. ChatGPT vs. DermGPT: Head-to-Head
- Research Letter: "Evaluating AI Models in Dermatology: ChatGPT vs DermGPT" (Patel et al, UC Irvine)
- Survey-based: Dermatology attendings and residents compared LLM responses to common derm questions.
- DermGPT preferred 48% of the time; ChatGPT 28%.
- Notably, DermGPT sometimes refuses to answer or refers to a dermatologist if the query exceeds safe boundaries, reducing risk of hallucination.
- ChatGPT's reference formatting was preferred, and it cited higher-impact journals.
- Quote:
“DERMGPT answers were preferred...but ChatGPT’s references were more frequently preferred...So basically there were more tight, to the point, better answers out of DermGPT.”
— Laura Ferris (24:40)
6. How Specialty-Specific AI Works
- Dr. Kmanger details DermGPT’s approach:
- Not just a rebadged ChatGPT—built as a “retrieval augmented generation model.”
- Curation: Initially pulled all PubMed articles, but curated down to what is relevant, practical, and high-quality for dermatologists.
- Multi-level agent checking system ("Derm Guardian") to minimize errors and prevent bad advice, akin to a panel of dermatologists checking one another’s work.
- Most common use cases: second opinions, nurse triage, prior authorization letters, and more.
- Quote:
“We have these models that check each other...probably one of the highest ways you can get to almost zero hallucinations…like having four attendings in a room checking each other.”
— Farah Kmanger (48:26)
7. Practical AI in the Clinic & What’s Next
-
Significant time saved on “pajama time” and administrative tasks.
- Up to a three-hour in-basket session reduced to 30 minutes with AI help (38:26).
-
True integration with EMRs remains a technological and institutional challenge.
- Current AI usually operates outside the EMR; full functionality requires seamless integration.
- Burnout in dermatology is rising due to growing demands; AI could meaningfully reduce this administrative load.
-
Looking Forward:
- Foundational LLMs (e.g., ChatGPT, Claude, Gemini) improve constantly, but specialty models add workflow and nuanced expertise.
- Integration, curation, and physician insights are where specialty AI will maintain an edge—even as general models improve.
Notable Quotes & Memorable Moments
-
On AI's friendliness:
“They’re created to just be super nice, like super customer servicey. They’re created for you to want to keep engaging with them.”
— Farah Kmanger (06:48) -
On medicine’s trust paradox:
“...People hold self-driving up to a standard of perfection, whereas the right standard is 1% better than humans... Doctors are worse.”
— Matt Zirwas (09:49) -
Derailing old fears:
“I think we’re still okay for just a little while though.”
— Farah Kmanger (08:06) -
Physician's new role:
“I do actually think derms are safe because we're so procedurally based. ...More of our time shifts to doing procedures. ...I think that's ten years out before that happens.”
— Matt Zirwas (11:23) -
On reference formatting wars:
“We should put some effort into making [DermGPT’s references] look pretty to...”
— Farah Kmanger (27:18)
Key Timestamps
- 03:26 – 04:54: ChatGPT-4 outperforms dermatologists in patient queries
- 06:00: Dr. Kmanger on AI adoption cycles and rapid uptake of generative AI in medicine
- 12:29: AI's ability to track and compare moles over time
- 17:16: “Digital twin” video avatars as the future of patient counseling
- 24:40 – 26:52: DermGPT vs. ChatGPT head-to-head results
- 30:17 – 36:09: Deep dive: How DermGPT’s technical and practical layer differs from general LLMs
- 38:26: Concrete time-savings with AI tools in real practice
- 41:04 – 44:37: The challenge of integrating AI directly with EMRs and the advantage of workflow-specific tuning
- 48:26: Multi-tier agent systems: maximizing accuracy and risk mitigation in AI answers
Episode Conclusion
The take-home: AI is already outperforming physicians in some aspects of communication and knowledge delivery, especially for routine queries. Both patients and doctors recognize these strengths but aren’t quite ready to give up the human touch. Specialty-focused AI models like DermGPT are carving their niche by curating medical knowledge, reducing administrative burdens, and integrating tailored workflows. While full replacement is still years away, AI will keep expanding in scope—reshaping dermatology and, potentially, redefining what it means to be a doctor in an increasingly digital world.
Memorable Closing Exchange:
“We’re all going to be sitting at home with our humanoid robots taking care of us and maybe doing podcasts so that we have something to do.”
— Matt Zirwas (14:31)
Additional Segment: AI Trivia (52:21 – 56:35)
A lighthearted round of trivia covers the Turing Test, AI in films (“Alien” and “Blade Runner”), with the hosts riffing on how fast conversation about AI has moved beyond simple benchmarks—marking the relentless acceleration of the field.
Finished Listening? The future isn’t coming for your derm job tomorrow, but it’s time to put AI to work for you—before it works instead of you.