Does AI Replace Labor (yet) ? Evidence from Machine Translation
Using data from one of the world's major freelancing platforms, we investigate the impact machine translation—an important application area for AI—has on labor. We use the introduction of neural network-based translation by Google (GNMT) as a natural experiment to examine AI affects the number of language translation transactions on a platform. The results suggest that GNMT replaced human labor. Specifically, we found a considerable drop in the number of language transactions on the platform for languages that could be translated using GNMT. The paper also shows that GNMT had a greater influence on standardized translation jobs than on creative ones. The findings suggest that, in the near and medium term, AI will likely reduce the scope of work performed by human labor.