Chomsky and the Two Cultures of Statistical Learning

At MIT's 150th birthday symposium in 2011, Noam Chomsky criticized machine learning researchers for using purely statistical methods to mimic linguistic behavior without understanding its underlying meaning, arguing that accurately modeling linguistic facts without identifying fundamental principles is merely "butterfly collecting" rather than true science. The essay responds to Chomsky's critique by arguing that engineering success demonstrates something is working correctly and should count as evidence of scientific success, and that accumulating facts through statistical models is consistent with how science has historically progressed, not a novel departure from it.