Kirk Borne's review of When Not to Use Deep Learning