The WHO Guidance for the Use of Large Multi-Modal Models in Health
Issued January 2024
Summary
Artificial intelligence (AI) holds great promise in healthcare, with Large Multi-Modal Models (LMMs) emerging alongside Large Language Models (LLMs) and generative AI chatbots. LMMs can process diverse data types like text, images, video, and sound, making them valuable for comprehensive medical analysis. To address the need for robust governance of AI in the health sector, the World Health Organization (WHO) released this new guidance document.
What is it about
LMMs are AI models that can process different data types (text, images, video, sound) at the same time. This makes them useful for analysing medical data which comes in these formats.
Who is it for
This guidance is directed towards developers, deployers, and governments who are involved with LLMs in countries that are members of the WHO.
Related guidelines
This guideline builds upon The WHO 2021 guidance which identified ethical challenges and risks of AI, alongside six consensus principles.
Key Takeaways
LMMs in healthcare present novel opportunities and risks for diagnosis and clinical care, patient-guided use, administrative tasks, medical and nursing education, and scientific research and drug development.
Governments hold a key responsibility to develop, implement and enforce appropriate laws, rules and regulations to protect individual rights and data, and to ensure responsible development, provision and deployment of LMMs.
Developers play a crucial role in establishing responsible practices for LMMs through their design and development. Key recommendations emphasize their responsibility to protect personal data, maintain accurate model performance, and transparently communicate dataset limitations.
Deployers take part in the governance of LMMs by avoiding the deployment of LMMs in inappropriate settings, communicating actual and potential risks of LLM use in specific settings and promoting accessibility to LMMs.
World Health Organization. (2024). Ethics and governance of artificial intelligence for health: Guidance on large multi-modal models. Chief Scientist and Science Division (SCI), Health Ethics & Governance (HEG). ISBN: 978-92-4-008475-9. Figure 3, p 18.
Why is this guidance important?
"This guidance is a compass to guide the development and equitable distribution of LLMs in specific to the health sector. It is unique because of its specific focus, is tailored for global applicability, encompasses diverse contexts, and it is crafted by a diverse expert group spanning disciplines and regions."
Prof. Dr. Effy Vayena
Co-chair of the WHO expert group on ethics and governance of AI for health