This month, the World Health Organization introduced new guidelines on the ethics and governance of large language models (LLMs) in the healthcare industry. The reaction from healthcare AI companies’ leaders have been predominantly positive.
WHO identified five main applications for LLMs in healthcare: diagnosis and clinical care, administrative tasks, education, drug research and development, and patient-guided learning. While recognizing their potential in enhancing global healthcare—such as assisting to alleviate clinical burnout and accelerate drug research—the organization warned of a tendency to “overstate and overestimate” the capabilities of AI. This, they argue, may lead to the adoption of “unproven products” which have not been rigorously tested for efficacy and safety. Addressing this issue, WHO stressed the need to move away from “technological solutionism,” a mindset that overemphasizes AI’s potential in solving deep-seated social, economic, and structural issues.
A noteworthy aspect of the guidelines is the involvement of various stakeholders in the design process of healthcare LLMs. They ought not to be solely created by engineers and scientists, WHO notes, but should also include healthcare providers, patients, and clinical researchers. The guidelines also stress the importance of transparency, inclusivity, and accountability in their design and deployment.
Piotr Orzechowski, CEO of Infermedica, lauded the guidelines as “a significant step” in encouraging responsible use of AI in healthcare settings. Jay Anders, CMO at healthcare software company Medicomp Systems, echoed these sentiments, advocating for external regulation for all healthcare AI. Other notable leaders, such as Michael Gao, CEO and co-founder of SmarterDx, shared that these guidelines serve as a balance between harnessing the benefits of AI while maintaining the provider-patient relationship’s integrity.
Contrarily, healthcare AI cannot forego innovation due to fears over the risks that LLMs pose, according to Gao. “A far greater risk is inaction in light of soaring healthcare costs,” he stated. Luz Eruz, the CTO of synthetic data company MDClone, noted the omission of synthetic data from the guidelines. According to Eruz, synthetic data, when combined with LLM’s, can enable researchers to parse and summarize vast amounts of patient data sans privacy issues, a topic ripe for regulatory attention.
With these guidelines, it appears WHO has initiated an important discourse on leveraging AI technologies like LLMs responsibly. As newer developments encroach into the healthcare sector, these principles laid down by WHO will act as a vital reference point for all stakeholders involved.
Further information could be found on the full article at Med City News.