Post date: Oct 25, 2017 7:17:0 PM
Not to complicate the matter, but the attached paper seems to support the need for algorithm testing standards/certifications. Not only is unintentional bias an issue, but the maliciously trained neural network is also a possibility.