This month, the World Health Organization (WHO) launched a tool it calls the Clever AI Resource Assistant for Health, or Sarah for miniature. However, the AI-based chatbot has already made some mistakes.
Chatbot provides users health information in eight different languages covering topics such as robust eating, mental health, cancer, heart disease and diabetes.
Developed by Modern Zealand company Soul Machines, Sarah can even apply facial recognition technology to provide more empathetic responses. Previous versions of Sarah have helped fight misinformation during the Covid-19 pandemic.
“SARAH gives us insight into how artificial intelligence can be used in the future to improve access to health information in a more interactive way,” said WHO Director-General Dr. Tedros Adhanom Ghebreyesus in a press release. “I call on the research community to support us continue to explore how this technology can reduce inequalities and support people access timely and reliable health information.”
While the WHO says the AI bot was trained based on the latest information from the agency and its trusted partners, Bloomberg Report that it does not provide up-to-date information on US-based medical advice and news events.
For example, when Bloomberg reporters asked Sarah whether the U.S. Food and Drug Administration had approved Lekanemab, an Alzheimer’s disease drug, the bot replied that the drug was still in clinical testing. In fact, the FDA approved the drug in 2023.
In another example, the bot was unable to provide details about the WHO’s report on hepatitis deaths until it was asked to check the agency’s website for updated statistics.
WHO did not immediately respond to a request for comment. On its AI bot landing page, the agency acknowledges that “responses may not always be correct because they are based on patterns and probabilities in available data.”
