Creation and Adoption of Large Language Models in Medicine

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Clinical Decision Support

Background:

• Growing interest in large language models (LLMs) for potential medical applications.
• Current LLM applications in medicine often lack specific medical training data and benefit verification.
• Risk of losing agency in shaping LLM integration if not actively involved.

Purpose of the Study:

• To emphasize the need for active engagement in shaping the use of LLMs in medicine.
• To highlight the importance of responsible development and deployment of LLM-powered medical tools.
• To advocate for a proactive approach in integrating artificial intelligence into healthcare.

Main Methods:

• Analysis of current trends in LLM adoption in medical tasks.
• Identification of gaps in LLM training and validation for healthcare.
• Framework proposal for active shaping of LLM development and deployment.

Main Results:

• LLM applications are increasingly used for medical tasks without adequate medical data training or benefit validation.
• Passive observation risks ceding control over how LLMs reshape medical practice.
• Active shaping requires provisioning relevant data, defining benefits, and real-world evaluation.

Conclusions:

• Active involvement is essential to harness the benefits of LLMs in medicine responsibly.
• Provisioning medical data, specifying benefits, and real-world testing are critical for safe LLM integration.
• Proactive shaping ensures LLMs serve to enhance, rather than disrupt, medical practice.