SANSA
Abstract: NuSCI team is developing a high-assurance Retrieval Augmented Generation architecture with uncertainty-quantified foundation models for healthcare where the AI responses are grounded to prior training or knowledge-base, free of hallucinations, and aligned with human experts. People are already querying foundation models (FMs) for medical help. However, they are prone to confabulations and hallucinations. How do we develop AI for healthcare that can be used by users of varying expertise from novice public to expert professionals? NuSCI team is addressing this challenge as the larger CHALO effort at SRI in the ARPA-H Digiheals program.
Susmit Jha
Technical Director, NuSCI
My research interests include artificial intelligence, formal methods, machine learning and dynamical systems.