re: #54 goddamnedfrank

Let’s at least try English. What appeals to the powers-that-be is that it removes humans from the decision making process. They can argue that the algorithm is unbiased, unlike humans. Except we all know that the algorithm has the biases the training data introduces, but people aren’t going to find those biases easily.

Now, modern ML techniques can actually find out what a network used to make its decision, as in which parts of the input were used and which were ignored.