This is a great response and I wish more scientists held publications accountable to their reporting. MIT Tech Review, Popular Science and Mechanics are read by a lot of people -- both literate in the your discipline and not, it's worthwhile to point the more nuanced view than what a 1000 word magazine article can point out. I've personally found reporting by ArsTechnica on recent security issues a good model. Wired occasionally comes through but has similar problems like this. Nautilus has its own biases but in general is good. I don't think bad reporting in these cases is necessarily out of malice but the lack of background on reporter's side on your field. And, perhaps sheer laziness. Remember when you procrastinated on writing that long overdue paper, I'd imagine reporters aren't immune to that too ;)<p>Convolutional Neural Nets are getting to a hype-level that I find pretty scary. We don't want another AI winter because people expect way too much too early without understanding the domain, only to lead to receeded interest in the field. Honest evaluation and crediting is invaluable to ensuring that.<p>Also --> What if there were a "rapgenius" for paper/article reviews where these comments from trusted sources can be curated and commented on? Not sure about viability, etc.. but could be interesting.