Paul Schmitt, Director of Engineering at Reynolds & Moore, joins Erik Reynolds to explore what the safety community still has not solved: how robots decide what to do. Paul brings decades of experience across Ford, iRobot, and autonomous trucking to a conversation about the behavior layer in physical AI.

They cover human-robot interaction research, including a study on how pedestrians actually respond to driverless vehicles, and why the answer had nothing to do with lights or sounds. Paul and Erik also dig into behavior hierarchies and the trolley problem applied to real robotics scenarios.

The conversation also gets into why current machine learning safety standards still read more like guidelines than engineering requirements. Paul shares details on his work as a co-author of "The ML FMEA in Action: Lessons from Applications of Machine Learning Safety," documenting how the open-sourced ML FMEA framework connects machine learning development decisions to system-level safety outcomes, with applications across autonomous vehicles, humanoid robotics, and beyond.

Read Paul's paper: The ML FMEA in Action: Lessons from Applications of Machine Learning Safety

Check out our work: www.reynolds-moore.com