From the Inside AI newsletter (register here, if interested)
MIT Technology Review analyzed the last 25 years of AI papers published in arXiv, one of the largest open-source databases for scientific papers.
By downloading the abstracts and tracking words, the team surfaced three distinct trends: a shift toward machine learning in the late 1990s and early 2000s, a movement toward neural networks beginning in 2010, and over the past few years a focus on reinforcement learning.
Paper available here, good read.