Machine learning relies on algorithms that “learn” from past examples, thereby relieving the programmer from having to write lines of code to deal with every eventuality.

Even when they’re not pure black boxes, machine-learning algorithms can sometimes become complex enough that even their creators don’t fully understand them.