Growth in the Age of Intelligent Machines: Identified Regimes and Refused Arrivals

Modern economies increasingly run on machines that learn, not just people. Does the theory of long-run growth survive when some agents are built rather than born, and can the record tell us anything new? This paper treats an economy's structure as a network recovered from how its growth rates move together. From such data the direction and broad speed of long-run growth can be read off, but finer detail cannot. The central finding is an asymmetry: whether growth is steady, slowing, or accelerating towards a runaway singularity can be identified, but the date of any such takeoff cannot. Machines change the data, not what the data can reveal. The paper proves these limits and illustrates them in a worked economy; it forecasts no growth rate and dates no singularity.