There was a famous case of a narrow purpose AI which was supposed to learn to identify pictures of tanks, it learned to identify pictures of cloudy days instead

There was a famous case of a narrow purpose AI which was supposed to learn to identify pictures of tanks, it learned to identify pictures of cloudy days instead. In another case, an AI designed to become an oscillator, became an amplifier instead, amplifying the 60Hz noise from the electrical grid. Kinda equivalent to a deep-frying juicer, isn't it? Both cases, and others like them, are quite well understood in hindsight, but were not predicted beforehand by the people involved.