Hippocratic AI has announced a partnership with NVIDIA to develop empathetic AI healthcare agents with super-low-latency conversational capabilities, termed "Empathy Inference." These AI agents are based on Hippocratic's safety-focused large language model (LLM), specifically designed for healthcare. The collaboration aims to enhance patient interactions, making them more seamless, personalised, and human-like.
Deployment and Impact of AI Healthcare Agents in the Medical Field
The AI agents are deployed by various healthcare entities, including health systems, payors, digital health companies, and pharmaceutical companies, to assist with low-risk, non-diagnostic patient-facing tasks over the phone. The goal is to mitigate staffing shortages and improve access, equity, and patient outcomes in healthcare. The partnership leverages NVIDIA's advanced technology stack, including its Avatar Cloud Engine suite and Riva models for automatic speech recognition and text-to-speech translation. This collaboration will focus on developing a super-low-latency inference platform to power real-time applications. In a demonstration video presented by NVIDIA, an AI "nurse" interacts with a patient recovering from an appendectomy. The AI provides aftercare guidance and addresses queries regarding the safety of specific antibiotics for a patient allergic to penicillin and diabetic. The interaction appears seamless and successful, typical of controlled demo scenarios. Hippocratic AI asserts that their research indicates AI agents often surpass human performance across various categories.
The Role of Speed and Safety in Enhancing Patient-AI Connections
Research by Hippocratic AI indicates that even small improvements in inference speed significantly enhance patients' emotional connection with AI healthcare agents. For instance, a decrease from 3 seconds to 2.2 seconds in inference time resulted in higher percentages of patients feeling that the AI cared about them and feeling comfortable confiding in the AI. Hippocratic AI's unique safety approach includes a primary model trained with evidence-based content, a constellation architecture with multiple specialist support models, and built-in guardrails to involve human supervision when necessary. This approach has enabled their generative AI healthcare agents to outperform GPT-4 and LLaMA2 70B Chat on safety benchmarks.
Hippocratic AI's Vision: Creating a Safe LLM for Global Healthcare Transformation
Hippocratic AI aims to create the safest Large Language Model (LLM) for healthcare, with the belief that such technology can significantly enhance healthcare accessibility and outcomes worldwide. Co-founded by CEO Munjal Shah, along with experts from various prestigious institutions and companies, including El Camino Health, Johns Hopkins, Washington University in St. Louis, Stanford, Google, and NVIDIA, the company has garnered $120M in funding from prominent investors like General Catalyst, Andreessen Horowitz, Premji Invest, and SV Angel. The company has engaged with over 40 beta partners to test its AI healthcare agents in various areas, including chronic care management, wellness coaching, health risk assessments, social determinants of health surveys, pre-operative outreach, and post-discharge follow-up.
Source: Hippocratic AI
Image Credit: iStock
