About three years ago, while ChatGPT and, later, the GPT-4 AI model were still in secret development at OpenAI, a group of Researchers at Microsoft Research started investigating the potential impact of this new technology on healthcare delivery and medical research. One result of this investigation was the publication of the book in early 2023, "The AI Revolution in Medicine: GPT-4 and Beyond," coauthored with Carey Goldberg and Isaac Kohane. But since the book was written while GPT-4 was still a secret to the world, it was largely a work of speculation. So now, with a bit more than two years of real-world observation and experience, what is actually happening, and what is likely to happen next in the world of healthcare? This talk will go into what we had predicted, and what we got right, what we missed, as a way to help us see a bit into the future.

The potential of artificial intelligence (AI) in medicine is vast, yet its translation into trusted, decision-grade evidence for clinical development and regulatory submission remains a fundamental challenge. We must move beyond theoretical promise to develop a practical blueprint for building an AI-powered evidence engine.

Drawing on real-world case studies, we will demonstrate how integrated, multimodal approaches are actively de-risking programs and accelerating regulatory decisions. We will explore the application of AI-driven disease progression models to enhance trial design, the development of purpose-built AI tools that democratize advanced analytics, and the strategic use of patient digital twins to guide development decisions. The discussion will extend to critical regulatory partnerships, including an FDA research collaboration to address RW disease outcome measurement, and the practical implications of emerging FDA guidance on AI.

The true evolution lies not in any single model, but in systematically integrating advanced patient modeling, causal reasoning, and pragmatic AI to create a new tier of evidence. This integrated approach ultimately builds clinical and regulatory confidence for us to deliver better, faster decisions for patients.