AI – WHO calls for strict supervision of chatbots in medicine

Faced with the craze for large multimodal language models (LMM), the WHO published a detailed report this Thursday, January 18, 2024. It includes more than 40 recommendations to strictly regulate the development and use of AI technologies in the medical and health field.

AI chatbots with human-like capabilities but not infallible

LMMs like ChatGPT are trained on huge text databases, allowing them to automatically generate new text consistently on almost any topic. They can understand complex questions, search for relevant information and formulate articulate responses.

Some recent AI models even show advanced reasoning abilities. For example, the Claude chatbot developed by Anthropic can refuse to provide dangerous health answers and recognize the limits of its knowledge.

But these chatbots do not really have a deep understanding of language and the world. Their responses may contain false or dangerous information, especially on sensitive topics like health. The risk of algorithmic bias is also high.

These limits of AI are not always obvious to users, hence the importance of strictly regulating the use of LMMs in the medical field according to the WHO.

Promising applications to improve diagnostics and patient relations

Despite the risks, the responsible use of AI chatbots could offer significant benefits for public health.

The WHO identifies several potential applications of LMMs in health:

• Answer written questions from patients to direct them to the right people, treatments and health services
• Provide personalized information to patients about their health status and treatment
• Analyze a patient’s medical data to help diagnose diseases
• Generate realistic patient simulations for training healthcare professionals
• Automate time-consuming administrative tasks to allow doctors to focus on care

Some hospitals and startups are already experimenting with AI chatbots for well-regulated uses such as making appointments or screening for certain disorders.

But the WHO warns against the dangers of hasty adoption without an adequate ethical and regulatory framework.

With these recommendations, WHO hopes to chart a path toward developing AI language models that are both useful and ethical in healthcare. The challenge is to take advantage of the potential of chatbots to improve the accessibility of care, the patient relationship, the training of healthcare professionals and the advancement of medical knowledge. But patient safety and well-being must remain the top priority. Hence the importance of proactive regulation and continuous risk assessment according to the WHO.