The Autonomous Laboratory

Abstract

A new class of systems can now design an experiment, run it with robotics, measure the result, and choose the next experiment — around the clock, with little human input. This analysis examines these self-driving laboratories across their intellectual lineage, hardware, and decision engines, engages the landmark campaigns including the failures, and argues that the binding constraint on machine-paced discovery is infrastructure rather than intelligence — making discovery throughput a strategic capability on the order of compute itself.