No common sense = catastrophic failure

As a once-upon-a-time machine learning expert, I know that Google recruited a batch of machine learning scientists. Very cool it seemed. But the reality ... machine learning doesn't have common sense no matter how good the machine learned criteria (see the story below). So, for mission critical elements like search, Google still prefers to use human tuned criteria to avoid catastrophic failure. Fascinating - and an interesting lesson in trust for automated learning - although the notion of capturing common sense has been a preoccupation of machine learning and AI for a long while (e.g. cyc).

(Google apparently does use Machine Learning for Spam detection and no doubt in other areas as well.)