WHO: AI in the Healthcare System Can Carry Risks and ‘Serious Errors’

Artificial intelligence (AI) currently selects donor organs for transplantation better than humans and responds more effectively to certain health questions from patients, but it also brings risks and ‘serious errors’, warned the World Health Organization.

The World Health Organization (WHO), based in Geneva, welcomes the enormous potential of artificial intelligence, which can be very useful, but warns that the introduction of AI treatment procedures into the healthcare system carries the danger of “completely incorrect” information.

– It is essential to assess the dangers of using LLMs (generated large language model tools) such as ChatGPT, as they can lead to serious errors in healthcare and undermine trust in AI – stated the WHO.

The UN health agency made this statement at a time when applications of artificial intelligence are rapidly gaining popularity. Concern arises from the fact that regular precautionary norms are not applied to LLMs.

Proponents of using AI believe that doctors will soon use AI chat systems to respond more quickly to patients’ health questions, and AI will also be used in diagnostics.

However, two chatbots, from Microsoft and Google, have shown that their responses are not entirely reliable when it comes to facts, and the WHO is concerned precisely because these responses seem reliable.

– LLMs produce answers that may appear authoritative and credible to the end user. However, these answers can be completely inaccurate or contain serious errors, especially when it comes to health-related responses – warns the WHO.

This information can mislead and produce information that seems very credible, which the public cannot distinguish from reliable information.

Therefore, the WHO suggests that before applying AI in healthcare, the profession should address these issues and measure the benefits and harms before approving its widespread use.