TFA: The results are intriguing—even a relatively simple neural network can be used to over-interpret an image, just like as children we enjoyed watching clouds and interpreting the random shapes. This network was trained mostly on images of animals, so naturally it tends to interpret shapes as animals.
Less intriguing: to consider that similar networks (especially once giving "recommendations" to unquestioning end users) might ascribe e.g. criminal propensity or lack of creditworthiness to the odd proverbial "innocent bystander" by over-amplifying distinctions they "think" to have learned.
The "Bad Blue sky" tank detector https://neil.fraser.name/writi... "might be apocryphal" (just like the Obstinate Lighthouse http://www.snopes.com/military...
http://www.heise.de/forum/heis...
I thought we'd all suffered enough advertising on our keyboards since 1995 already?
Thankfully the glorious Model M remains unbreakable and unaffected...
Fast, cheap, good: pick two.