Translating AI into Practice
• Neil D. Lawrence
Abstract
Translating machine learning from research prototypes into robust, reliable systems is one of the greatest challenges facing industry today. This talk explores the practical hurdles of deploying AI in real-world environments, drawing from over 25 years of experience in machine learning systems design at Amazon and insights from “The Atomic Human.”
We’ll examine the interface between machine learning and systems research, exploring how traditional software engineering practices must evolve to handle the unique challenges of data-dependent systems. The talk will cover deployment challenges, intellectual debt in ML systems, and the importance of continuous monitoring in production environments.
Central to the discussion is the human perspective: how do we build AI systems that complement rather than replace human expertise, particularly in critical domains like healthcare? We’ll explore trust, autonomy, and the essential role of human oversight in ensuring AI serves society’s needs while maintaining the reliability and safety standards that sectors like pharmaceuticals demand.