It's also important to note that the algorithms today cannot be audited seriously. We know, for example, that many municipalities have been caught installing red light cameras and then adjusting the yellow light timing to maximize revenue rather than maximize commuter safety. We know that some laws, particularly drug laws, have consequences designed to punish "undesirable groups" (minorities) rather than make streets safer or protect the lives of citizens. So having pseudo-AI that is also a black box isn't necessarily worse than what we have today. Because as of today we can't really audit the algorithm itself, only its output.