There are different ways of estimating automation risk. This paper presents a new method to predict the impacts of a technology on occupations using the overlap between the text of job task descriptions and the text of patents. This is used to construct a measure of the exposure of tasks to automation. The author first applies the method to historical cases such as software and industrial robots, establishes that occupations measured as highly exposed to previous automation technologies saw declines in employment and wages over the relevant periods. The author then uses the fitted parameters from the case studies to predict the impacts of artificial intelligence.
The findings suggest that new AI applications, as described in patents, will replace high-skilled tasks. This could alter structures of inequality.
Under the assumption that the historical pattern of long-run substitution will continue, the author estimates that AI will reduce 90:10 wage inequality, but will not affect the top 1%.