At the recent NVIDIA GTC conference, Munjal Shah, the founder of Hippocratic AI, took the stage to showcase one of the first real-world deployments of large language models (LLMs) in the healthcare sector. Hippocratic AI, a startup backed by over $120 million in funding from investors such as General Catalyst, (a16z) Bio + Health, and Premji Invest, has been working for more than a year to create “empathetic” artificial intelligence agents capable of engaging in nuanced voice conversations with patients.
Munjal Shah’s vision for Hippocratic AI is to safely deploy these AI agents to perform a wide range of non-diagnostic tasks, which can help improve patient outcomes and alleviate the stresses faced by human clinicians due to staffing shortages. These AI agents, designed with a human-centric approach, will provide preoperative or postoperative guidance, gently encouraging patients to adhere to their care plans, and answering questions about medications, all while conveying a friendly, familiar, and caring demeanor.
“With generative AI, patient interactions can be seamless, personalized, and conversational,” Munjal Shah said in a recent statement on the company’s partnership with NVIDIA to reduce latency and improve the speed of inference, which is crucial for achieving the desired impact. “Every half-second of reduced latency,” Shah added, “increased patients’ sense of emotional connection by up to 10%.”
Munjal Shah emphasized that NVIDIA’s powerful AI chips are critical in achieving the speed and fluidity for these AI agents to have natural, human-like conversations. Hippocratic AI has termed this low-latency AI an “empathy inference engine.”
In a recent interview, Munjal Shah acknowledged the shortcomings of older interactive voice response (IVR) systems, which often struggled with comprehension and required users to speak in a particular way. However, the advancements in LLMs have transformed this landscape, with Hippocratic AI agents boasting significantly higher comprehension and speech synthesis capabilities.
Hippocratic AI’s approach to training its LLMs strongly emphasizes safety considerations, aligning with the company’s motto of “do no harm.” This involves creating a constellation of LLMs trained solely on authoritative, evidence-based medical sources, subjecting them to rigorous reinforcement learning and testing by human medical professionals. Only once these professionals are satisfied with the safety and effectiveness of the LLMs will they be made commercially available.
Munjal Shah stressed that Hippocratic AI’s agents are designed solely for straightforward, non-diagnostic workflows that can reduce routine follow-up or administrative burdens on clinicians and other staff rather than providing high-stakes medical advice. Some initial use cases include preoperative and postoperative outreach to guide patients, onboarding patients to new medications, and reminding patients about adherence routines.
The opportunity for Hippocratic AI’s ’empathetic agents’ is substantial, given the projected shortage of 275,000 additional nurses needed in the United States from 2020 to 2030, as the population ages and care needs intensify. The potential of automating these time-consuming tasks can help reduce burnout among overworked nurses while allowing them to focus on more complex and urgent care, emotional support, counseling, and answering diagnostic and treatment questions, offering a hopeful future for healthcare.
Munjal Shah’s vision for Hippocratic AI is to augment human healthcare staff by focusing on the touch points that can increase patient access to care, make many interactions more convenient, and improve outcomes, which healthcare staff often lack the time to address. With the collaborative support of NVIDIA’s cutting-edge technology, Hippocratic AI is poised to make a significant impact on the healthcare industry by leveraging the power of LLMs to enhance patient experiences and alleviate the burden on human clinicians.
