Inventions depend on skills, experience, and information exchange. Information is shared among individuals and organizations both intentionally and unintentionally. Unintentional flows of knowledge, or knowledge spillovers, are viewed as an integral element of technological progress. However, little is known about the overall patterns of knowledge flows across technology sectors or over long periods of time. This paper explores whether it is possible to identify “invention machines” – technologies that help create new inventions in a wide range of other sectors – and whether shifts in the patterns of knowledge flows can predict future technological change. In the spirit of big data we analyze the entire PatStat database of 90 million published patents from 160 patent offices over a century of invention and exploit variation within and across countries and technology fields over time. The direction and intensity of knowledge spillovers measured from prior-art citations highlight the transition from mechanical to electrical instruments, especially industrial control systems, and the rise of information and communication technologies as “invention machines” after 1970. Most recently, the rapidly increasing impact of digital communications on other fields may herald the emergence of cloud computing and the industrial internet as the new dominant industrial paradigm.