Skip to content

Large language models in medicine

Authors: Arun James Thirunavukarasu; Darren Shu Jeng Ting; Kabilan Elangovan; Laura Gutierrez; Ting Fang Tan; Daniel Shu Wei Ting
Year: 2023
Venue: Nature Medicine

type: survey collection: sok-references status: auto

Links: DOI

Summary

Auto-extracted summary (unreviewed)

A clinician-oriented review of large language models in medicine, using ChatGPT (GPT-3.5/GPT-4) as the running example. It explains how LLM chatbots are developed (pretraining, fine-tuning, RLHF), surveys clinical, educational and research applications reported so far, and discusses limitations such as inaccuracy, bias, privacy and the need for validation through pragmatic clinical trials. The authors conclude that autonomous deployment is not yet feasible and clinicians remain responsible for care.

Abstract

Large language models (LLMs) can produce useful responses to free-text queries without being specifically trained with many examples of the task in question, causing excitement and concern about artificial intelligence (AI) systems being implemented in healthcare settings. ChatGPT is a generative AI chatbot produced through sophisticated finetuning of an LLM (GPT-3.5 or GPT-4), and other tools are emerging through similar developmental processes. Here, we outline how LLMs such as ChatGPT are developed, show how LLMs are being leveraged in clinical settings, and consider the strengths and limitations of LLMs with regards to their potential applications which may improve the efficiency and effectiveness of clinical, educational, and research work in medicine. LLM chatbots have already been deployed in a range of biomedical contexts, with impressive but mixed results. This review acts as a primer for interested clinicians, who will determine if and how LLM technology is used in healthcare for the benefit of patients and practitioners.
Source: pdf

Extracted details

Field Value Source Confidence
contribution primer on how LLM chatbots such as ChatGPT are developed llm 0.85
contribution overview of clinical, educational and research applications of LLMs in medicine llm 0.85
contribution discussion of strengths, limitations and risks of LLMs in healthcare llm 0.85
contribution recommendations for validation via pragmatic clinical trials with transparent reporting llm 0.85
models GPT-3 llm 0.85
models GPT-3.5 / ChatGPT llm 0.85
models GPT-4 llm 0.85
models Med-PaLM 2 llm 0.85
models Bing AI chatbot llm 0.85
tasks medical question answering llm 0.85
tasks clinical decision support llm 0.85
tasks medical education llm 0.85
tasks clinical documentation and administration llm 0.85
tasks research writing llm 0.85
interaction other llm 0.85
limitations LLM responses can be inaccurate or fabricated, with poorly calibrated uncertainty llm 0.85
limitations confidential architecture, training data and alignment of GPT-4 limit scrutiny llm 0.85
limitations ethical, privacy and bias concerns; autonomous clinical deployment not currently feasible llm 0.85
limitations clinical benefit must be shown in validated, pragmatic trials llm 0.85
llm relevance low llm 0.85

SoK reference [1]

A. J. Thirunavukarasu, D. S. J. Ting, K. Elangovan, L. Gutierrez, T. F. Tan, and D. S. W. Ting, “Large language models in medicine,” Nature medicine, vol. 29, no. 8, pp. 1930–1940, 2023.

Listed on the reference page of the site.

Keyword mentions

Counts of vocabulary terms in the full text (first page of occurrence in brackets).

  • model: GPT 79 [p.1], Transformer 11 [p.2], LLaMA 7 [p.3], BERT 3 [p.3], ViT 2 [p.9]
  • dataset: COVID 2 [p.6]
  • technique: Fine-tuning 6 [p.3]
  • task: Medical / healthcare 125 [p.1], Language modeling / generation 5 [p.3]
  • topic: Large language models 119 [p.1], Benchmark 7 [p.2]
  • GPT-4 System Card misfiled as this paper — the PDF was originally filed under this paper

Files

Path (under papers/) Collection Pages Size SHA-256 Note
references/sok-references/001_large-language-models-in-medicine.pdf sok-references 17 1.4 MB 8cf54ca38468

Notes

Local PDF is the author's submitted manuscript (CC BY 4.0, Apollo repository https://doi.org/10.17863/CAM.96638) converted from DOCX; its three EMF figures are not rendered. Published version: Nature Medicine 29, 1930-1940 (2023).


Database key: large-language-models-in-medicine · created 2026-11-05 · updated 2026-11-05 · edit with python3 tools/paperdb.py set large-language-models-in-medicine COLUMN VALUE