The World Health Organization said it has expanded its generative artificial intelligence assistant, first launched during the COVID-19 pandemic as Florence, to help people everywhere live healthier lives.
S.A.R.A.H, or Sarah, was created using an AI training approach based on modeling interactive human systems to deliver more personalized and engaging messages.
WHY IT MATTERS
WHO said in its announcement Tuesday that Sarah has the expertise to help prevent cancer, heart disease, lung disease, diabetes and other causes of high mortality rates.
To create the digital health promoter, WHO partnered with New Zealand-based Soul Machines, an interactive AI development firm.
The company published a Medium post on Tuesday to explain how it designs its biological AI with sensory, motor, attention and perception to create digital workforce solutions.
We reached out to Greg Cross, CEO of Soul Machines, to ask about algorithm development, how the company ensures bias and discrimination are minimized, and the maintenance of Sarah’s data hygiene over time.
In addition to using large language models that invest heavily into data training sets that minimize bias, he said that Sarah’s data is optimized for global diversity.
“There is additional retrieval-augmented generation training that the WHO applies for continual optimization across a very diverse set of geographies, languages and cultures training that will improve over time,” Cross told Healthcare IT News by email.
However, Sarah is expected to get better at interactions over time because multimodal cognitive modeling is used to simulate human behavior through the company’s patented “digital brain technology.”
“And most importantly, with the cognitive modeling approach to our patented technology, biological AI, our digital people learn from the data like the way humans do – by interaction and experience,” he explained.
“This helps generate more accurate, relevant experiences and outcomes for this broader, diverse audience.”
The goal for healthcare AI tools like Sarah is to fill healthcare’s gaps, not replace its “heroes,” the company said in the blog. Its lifelike countenance is essential to delivering personalized, interactive support, including wellbeing tools.
“This is particularly crucial for geographically isolated populations or those facing chronic illness,” Soul Machines said.
Meanwhile, WHO described Sarah’s use of generative AI to deliver health messages based on the sources used to train the LLMs.
“The answers may not always be accurate because they are based on patterns and probabilities in the available data,” the organization said on its Sarah website.
The global organizing body said it encourages researchers to look into how Sarah might improve health equity.
THE LARGER TREND
With an overburdened healthcare system, experimenting with genAI has led to an emerging ecosystem of tools that does everything from finding information in electronic health records to documenting clinical notes and automating administrative tasks to providing clinical decision support.
But as AI is susceptible to errors, including unintended outputs, the deployment of AI as a substitute for medical advice and at the point of care is in its nascency.
While AI may boost CDS in emergency and other settings, key decisions can be fraught with high variability, bias and limited prognostic validity, according to Scott Levin, senior director of research and innovation at Beckman Coulter, and professor in emergency medicine at Johns Hopkins University School of Medicine.
“It is vital for healthcare to have a framework for how AI tools address challenges, are developed, implemented and evaluated for impact,” he told Healthcare IT News ahead of his HIMSS24 educational session.
“This includes studying how clinicians interact with these tools and how it may change their decision-making behavior,” he said.
ON THE RECORD
“The future of health is digital, and supporting countries to harness the power of digital technologies for health is a priority for WHO,” Dr. Tedros Adhanom Ghebreyesus, WHO Director-General, said in a statement.
“S.A.R.A.H. gives us a glimpse of how artificial intelligence could be used in the future to improve access to health information in a more interactive way.”
Andrea Fox is senior editor of Healthcare IT News.
Email: afox@himss.org
Healthcare IT News is a HIMSS Media publication.