The language inside patent documents can predict a massive technological breakthrough 20 years before it actually happens.
Predictive patterns are hidden in the way inventors describe their ideas long before the industry recognizes a revolution. Large language models can spot subtle shifts in vocabulary and technical framing that signal the birth of a new field. People usually think of innovation as a sudden eureka moment that takes the world by surprise. These results show that the seeds of the future are planted in plain sight decades in advance. Governments and investors could use this to forecast the next industrial shift simply by reading the fine print of existing filings.
Anticipating Innovation Using Large Language Models
arXiv · 2605.04875
Forecasting innovation, intended as the emergence of new technological combinations, is a fundamental challenge for science and policy. We show that forthcoming combinations leave an early trace in the collective language of patents, with predictive signals detectable even decades in advance. We show that signal is not attributable to any single inventor, but emerges as a collective shift in how technologies are described across thousands of patents. To this end, we introduce TechToken, a transf