Why doctors should care
- Generative AI has moved from hype to bedside utility.
- Ambient “copilot” apps draft encounter notes.
- Large language models (LLMs) summarise new studies in seconds.
- Early adopters save hours each week and report less burnout.
Five core concepts, clinical parallels
- Neural networks → Repeated pattern-finding, like residents running endless simulations.
Example: Abridge and Nuance DAX learn note structure from thousands of visits. - Prompt engineering → Give precise orders, just as in the ED.
Add specialty context, patient age, and units for first-pass accuracy. - LLMs → Word prediction engines that now read images too.
GPT-4o produces plain-language study briefs; Med-PaLM 2 stays safe on citations. - Image generators → Sentence-to-slide tools for education.
Create rare-disease X-rays for resident drills without exposing PHI. - Video generators → Extend images through time.
Eight-second clips teach inhaler technique in clinic waiting rooms.
A prompt formula that works
Role → Task → Context → Constraints → Expected output
Example prompt: “You are a board-certified cardiology scribe. Draft a SOAP note for a 62-year-old male with new-onset atrial fibrillation. Include ICD-10 codes, no billing modifiers, plain text, under 250 words.”
Keep related words close together so the model links them correctly.
Quick pilot projects (about one hour each)
- Ambient notes: Test a free scribe in three low-acuity visits and measure edit time.
- Literature summaries: Ask an LLM for three bedside takeaways from this week’s NEJM article.
- Synthetic imaging: Generate 20 pneumothorax variants for the next M&M.
- Micro-videos: Script an eight-second Veo clip on the DASH diet and load it into the portal.
- Voice clones: Produce a multilingual discharge reminder with 11 Labs and test patient recall.
Guardrails to keep in place
- Verify all citations and calculations.
- De-identify data before upload and use BAA-compliant vendors.
- Audit bias across age, sex, and language groups.
- Follow current FDA guidance on AI-enabled devices.
What comes next?
Context windows already hold entire textbooks. Tomorrow’s copilot could scan your complete EMR and the latest guidelines before suggesting management. Speed will keep rising, but clinical judgment remains essential.
The bottom line is to train AI like a junior colleague. Give clear instructions, double-check its work, and let it handle routine tasks so you can focus on the art of medicine.
Harvey Castro is a physician, health care consultant, and serial entrepreneur with extensive experience in the health care industry. He can be reached on his website, harveycastromd.info, Twitter @HarveycastroMD, Facebook, Instagram, and YouTube. He is the author of Bing Copilot and Other LLM: Revolutionizing Healthcare With AI, Solving Infamous Cases with Artificial Intelligence, The AI-Driven Entrepreneur: Unlocking Entrepreneurial Success with Artificial Intelligence Strategies and Insights, ChatGPT and Healthcare: The Key To The New Future of Medicine, ChatGPT and Healthcare: Unlocking The Potential Of Patient Empowerment, Revolutionize Your Health and Fitness with ChatGPT’s Modern Weight Loss Hacks, Success Reinvention, and Apple Vision Healthcare Pioneers: A Community for Professionals & Patients.
