[Hippocrates and language models - Primum Non Nocere - First do no harm]

Area of Science:

• Medical Ethics
• Artificial Intelligence in Healthcare
• Natural Language Processing

Background:

• The Hippocratic Oath, a foundational ethical code for physicians, mandates the principle of 'First do no harm.'
• Rapid advancements in artificial intelligence (AI) and large language models (LLMs) present novel challenges and opportunities in medical practice.
• Integrating AI into healthcare requires careful consideration of ethical frameworks to ensure patient safety and well-being.

Purpose of the Study:

• To analyze the ethical considerations of applying large language models (LLMs) in clinical settings.
• To evaluate potential risks and benefits associated with AI-driven medical tools.
• To propose guidelines for the responsible implementation of LLMs in healthcare, aligning with the 'First do no harm' principle.

Main Methods:

• Systematic literature review of AI ethics in medicine.
• Analysis of case studies involving LLM applications in healthcare.
• Ethical framework development based on established medical principles.

Main Results:

• Identified potential biases in LLM outputs that could lead to diagnostic errors.
• Highlighted concerns regarding data privacy and security with AI implementation.
• Recognized the need for robust validation and oversight mechanisms for medical AI.

Conclusions:

• Adherence to the 'First do no harm' principle is paramount when deploying LLMs in medicine.
• Multidisciplinary collaboration is essential for developing safe and effective AI healthcare solutions.
• Continuous ethical evaluation and adaptation are necessary as AI technology evolves.