to train the network on New York Times articles, with the goal of teaching it to identify authorship. Sutskever concedes this could likely form the basis for plagiarism software one day, provided it functions well enough.