SIG: Real-World Evidence (RWE) Collaborative

 View Only
Back to discussions
Expand all | Collapse all

Guardrails for LLMs in PV and AI-generated podcast

  • 1.  Guardrails for LLMs in PV and AI-generated podcast

    Posted 10-02-2024 13:11
      |   view attached

    Hi everyone,

    It was wonderful seeing everyone in Berlin! As a follow-up to our discussion and as promised, sharing with you the attached pre-print from one of our colleagues (Darmendra Ramcharran, GSK).

    The paper outlines the development and application of a series of guardrails aimed at mitigating risks and improving the reliability of LLM outputs particularly in the realm of pharmacovigilance though these same guardrails could be considered for all types of RWE research.

    Also, if you are interested, here is an AI-generated 10 minute podcast developed by Google's NotebookLM (LINK) that talks about the paper (@Darmendra, you can tell me how accurate this is!).

    Thanks everyone and talk soon!



    ------------------------------
    John Diaz-Decaro
    Latent Virus, Epidemiology Lead
    ModernaTX, Inc.
    email: john.diaz-decaro@modernatx.com

    Vice Chair
    Digital Epidemiology Special Interest Group
    ------------------------------