Munjal Shah Bets on Health Care LLMs for ‘Super Staffing

Serial tech entrepreneur Munjal Shah sees major potential in using large language models (LLMs) to help address the global healthcare staffing crisis. However, he believes it must be significant thoughtfully, avoiding high-risk applications like diagnosis. Instead, his new startup, H, Hippocratic AI, focuses on leveraging LLMs as virtual assistants for non-diagnostic tasks.

Shah has an extensive background in AI, noting he built his first small neural network in 1992. However, he says today’s generative models represent a true breakthrough. Unlike past “classifier” AI that labels data, generative AI can create original content. Munjal Shah believes generative models still need to be appreciated despite the hype.

He warns that having LLMs act as virtual doctors could “actually kill somebody,” as they sometimes “hallucinate” false information. However, Shah identified major opportunities for using LLMs in supportive, non-diagnostic roles. This could help address the over 15 million global shortage of nurses, significant patient navigators, and other critical staff.

Hippocratic AI’s LLM aims to converse naturally with patients to assist with medication adherence, appointment coordination, food access, and overall chronic care support. Shah explains there needs to be more staff to provide this time-intensive care to all patients. But infinitely scalable LLMs could take on this “super staffing” role.

Shah believes effectively training the LLM nonmedical language is crucial. He highlights that patients need to engage with it, so they must communicate knowledgeably and empathetically. Hippocratic AI refines its model by getting feedback from real medical professionals. The goal is to be an “autonomous agent” providing health services worldwide.

Rather than reckless deployment, Shah advocates thoughtfully tapping LLMs’ communication strengths while avoiding high-risk diagnostic applications prone to hallucinated outputs. By focusing virtual assistants on supporting patients’ chronic care needs, Hippocratic AI aims to demonstrate LLMs’ promise for mitigating the healthcare staffing crisis through scalable, personalized interactions.

With his long track record launching startups, Shah sees a revolutionary opportunity in generative AI. However, success will require carefully defining the technology’s role to complement human providers, not staying within capabilities. Hippocratic AI represents his belief that LLMs can drive better patient engagement and outcomes when narrowly targeted at expanding the healthcare workforce.