Emilio Ferrara at the University of Southern California and his colleagues have trained an AI to detect bots on Twitter based on differences in patterns of activity between real and fake accounts.