Title

AWS re:Invent 2023 - Generative AI in medicine: Improving patient & provider experiences (AIM109)

Summary

The panel included leaders from Gilead, Eli Lilly, Bristol Myers Squibb, and AWS discussing the role of Generative AI (Gen AI) in the pharmaceutical industry.
The consensus was that Gen AI is both real and overhyped, with significant potential to transform the industry.
Key focus areas for Gen AI include drug discovery, clinical trials, patient experiences, and operational efficiency.
Gilead has adopted AWS as its strategic cloud service provider and is focusing on patient outcomes, aiming to bring transformative therapies to patients by 2030.
Eli Lilly aims to reimagine operations in an AI-powered world, accelerating drug discovery and clinical trials.
Bristol Myers Squibb is navigating upcoming challenges with the help of AI, focusing on innovation, commercial success, and efficient manufacturing.
AWS sees a shift from a hype cycle to a budget cycle, indicating that Gen AI will become more financially grounded.
Concerns include managing costs, especially with the increasing use of GPUs, and the potential legal issues surrounding IP and data protection.
Companies are adopting various strategies for implementing Gen AI, including buying, building, partnering, and using open-source models.

Data Foundation: A strong data foundation is crucial for Gen AI success, particularly in managing unstructured data and ensuring data is discoverable and usable.
Operationalizing AI: Companies are learning to operationalize AI, focusing on MLOps and LLM ops to manage the lifecycle and costs of AI applications.
Cost Management: There is a need for cost-aware architecture and cloud financial operations (FinOps) to manage the increasing costs associated with cloud and GPU usage.
Talent and Change Management: Upskilling existing talent to work with AI and managing the change within organizations are significant challenges.
Regulatory Concerns: The potential for regulatory changes and the need for responsible AI policies are concerns as the technology evolves.
IP and Attribution: There is a risk associated with IP attribution when using models to generate new IP, which could lead to future legal disputes.
Model Selection: Companies are considering the use of various models and the implications of model bias, as well as the need for models specific to their use cases.
ROI and Efficacy: Not all AI ideas are worth pursuing; companies are learning to focus on those with a clear ROI and data to support efficacy and accuracy.
Multi-Cloud and Multi-Model: The landscape is becoming more dynamic with the need to consider multi-cloud strategies and the use of multiple AI models.