CLOSE
Loading...
12° Nicosia,
22 December, 2024
 
Home  /  News

WHO introduces SARAH

Virtual health worker SARAH raises concerns over accuracy amid global staffing shortages

Newsroom

The World Health Organization (WHO) is venturing into the realm of artificial intelligence (AI) to disseminate health information worldwide through a human-like avatar.

Named SARAH, which stands for Smart AI Resource Assistant for Health, this virtual health worker is accessible 24/7 in eight languages to discuss topics like mental health, tobacco use, and healthy eating. SARAH represents the WHO's effort to leverage technology to educate people and address staffing shortages amid a global healthcare workforce crisis.

However, the WHO cautions on its website that SARAH's responses "may not always be accurate" as her knowledge is not up-to-date with the latest data. Consequently, the AI-powered robot sometimes provides erroneous answers, referred to as hallucinations in AI models, which could spread misinformation about public health.

SARAH has limitations too. Programmed to stay within the WHO's scope, SARAH avoids discussing topics beyond its purview, such as specific drugs. Instead, it often redirects users to the WHO website or advises consulting a healthcare provider.

Despite its shortcomings, the WHO envisions SARAH as a tool to collaborate with researchers and governments to enhance public health efforts. The agency seeks input from countries to refine the bot and deploy it in health emergencies. Nonetheless, the WHO emphasizes that SARAH is a work in progress and cannot replace interactions with trained healthcare professionals.

Trained on OpenAI's ChatGPT 3.5, SARAH lacks real-time medical information beyond September 2021. Consequently, she may provide outdated responses to queries. Instances of inaccuracies include misinformation about drug approvals and trends in disease mortality rates.

While AI holds promise in advancing healthcare, SARAH's early-stage development underscores the need for caution. Studies have shown that AI models like ChatGPT can produce correct answers most of the time but may also make significant errors, highlighting the importance of ongoing refinement and validation.

Source: moneyreview.gr, Bloomberg

TAGS
Cyprus  |  health  |  World

News: Latest Articles

X