As I've pointed our elsewhere in the thread, I think a lot of people aren't distinguishing between an explicit algorithm, and a machine learning algorithm. I think people in this thread are looking at algorithms as a black box, where you put data in, something incomprehensible happens, and then police go and arrest people. When you have machine learning, it's a non-deterministic process where even the programmer who built the system can't work it backward and say 'this person was arrested because of these inputs to the system.' But there are tons of algorithms that could be developed where the programmer can tell you EXACTLY which inputs lead to a particular result, and the transparency of these algorithms could vastly exceed the transparency of machine learning, and even exceed the transparency of our current human-driven system.