There are applications of machine learning that are well scoped, well tested, and involve appropriate training data such that they deserve their place among the tools we use on a regular basis. These include such everyday things as spell-checkers (no longer simple dictionary look-ups, but able to flag real words used incorrectly) and other more specialized technologies like image processing used by radiologists to determine which parts of a scan or X-ray require the most scrutiny. But in the cacophony of marketing and startup pitches, these sensible use cases are swamped by promises of machines that can effectively do magic, leading users to rely on them for information, decision-making, or cost savings—often to their detriment or to the detriment of others.