Large Language Models & Generative AI in Medicine

Large language models represent the most rapid adoption wave in medical AI history. Within months of GPT-4's release, every major health system and pharma company was evaluating LLM applications. The technology's versatility — understanding context, generating text, reasoning about complex scenarios, and following nuanced instructions — makes it applicable across clinical, operational, and research domains in ways that task-specific AI models cannot match. Medical-specific LLMs (Med-PaLM, BioGPT, ClinicalCamel) are tuned for healthcare accuracy, while general LLMs are being deployed through healthcare-specific platforms with guardrails.

Ambient clinical documentation is the killer application for LLMs in healthcare. Abridge and Nuance DAX Copilot use LLMs to listen to physician-patient conversations and generate structured clinical notes — understanding medical context, clinical reasoning, and documentation requirements. Abridge has been named Best in KLAS for two consecutive years, deployed at Sharp HealthCare, MaineHealth, and BILH. Merck's GPTeal initiative applies generative AI across pharmaceutical R&D. Bristol Myers Squibb partnered with Accenture to deploy GenAI in drug development. Epic, the dominant EHR vendor, is integrating LLM capabilities directly into clinical workflows used by over 250 million patients.

Generative AI is also transforming medical education, patient communication, and administrative operations. LLMs generate patient-friendly discharge instructions, translate complex medical information into accessible language, and draft prior authorization appeals with clinical justification. In research, generative AI assists with literature review, protocol development, and manuscript preparation. Drug design uses generative models to propose novel molecular structures and protein sequences. The primary challenges are hallucination (generating plausible but incorrect medical information), reliability (consistent performance across edge cases), and governance (ensuring appropriate oversight for clinical applications). Healthcare organizations are establishing AI governance frameworks that distinguish high-risk clinical LLM uses from lower-risk administrative applications, applying proportionate oversight to each.