The importance of "GOFAI" techniques was one of the themes in my talk at All Things Open last year. I wanted to cover the "state of the art" in open source machine learning tools, but realized that that just promoted FDD - Fad Driven Development, so I spent a modest amount of time during the talk going over some of the older ideas and pointing out how they still had relevance, and suggesting that the Next Big Thing might involve some hybridization of GOFAI techniques and probabilistic / machine learning ideas (including deep learning). So I guess I'd basically agree with the author here. The only thing I wish he/she would have done, is add some links to papers or whatever, highlighting what they consider to be some of the important new results in AI.
There's definitely still plenty of work ongoing that falls under the "AI" rubric, see the cs.AI category on Arxiv.org for evidence of that. Of course some things get cross-posted with the ML category as well, so it's not all "GOFAI".