Snapshot
- Building a Dedicated Team for Generative AI
- Identifying Key Use Cases for Generative AI in Healthcare
- Running Proof of Concept and Risk Assessment
- Choosing the Right Commercial Model and Specific LLM
- Establishing a Data Strategy for Specialized Use Cases
Overview
Building a Dedicated Team for Generative AI
Silva emphasizes the importance of dedicating a team with the right skills to understand generative AI advancements and the evolving tooling and vendor landscape in healthcare.
Identifying Key Use Cases for Generative AI in Healthcare
Mocingbird identified top use cases for generative AI by analyzing their business goals and current workflows, both internally and for customer interactions.
Running Proof of Concept and Risk Assessment
They conducted a proof of concept to test feasibility, focusing on the top use case. This involved considering risks such as AI hallucinations, assessing data availability and quality, and evaluating security implications.
Choosing the Right Commercial Model and Specific LLM
Given the risks around variance and accuracy, Mocingbird decided to use a domain-specific large language model (LLM) like Med-PaLM 2, which is trained specifically on medical data.
Establishing a Data Strategy for Specialized Use Cases
Finally, they defined outcome goals and accuracy metrics, creating a golden data set using proprietary data to fine-tune the chosen model for their specialized healthcare use case.
