Is There a Sense of Cleverness in the Formulation?

A lot of that work around 2010-2011-2012 was happening sort of first in computer vision, first on image classification. Then fairly quickly people got interested in when we apply analogous ideas and some of the same technologies to language tasks. It's this weird story where the trend has been towards less and less cleverness, more and more toward kind of do the simple thing, do the thing that had been kind of that somebody wrote down 20 or 30 or 40 years ago and just do it a lot and get the engineering right.