As AI becomes the go-to tool of the online world, companies are reporting only their successes, while the public and government regulators are kept in the dark as to how often these algorithms make mistakes and, most importantly, if they are right more often than they are wrong.