[xg15]

> In trying to apply a computer to a task that humans do, we often discover that it doesn't work. One common problem is that humans are able to deal with fuzzy concepts but computers are not -- they need precise representations and it is hard to represent a fuzzy concept in a precise way. However, if we look closer at such tasks, we often discover that the weakness actually lies not in the computer but in ourselves -- we didn't understand what we were doing in the first place.

This was still published before the latest iteration to evade the insight that your question is ill-posed took hold: AI magic!

Why go through the trouble to ask yourself if your classification system makes sense if you can just draw up some random examples and train a classifier? And if the classifier happens to perform poorly, it's clearly the fault of the algorithm and not that your classifications are completely arbitrary.

Modern algorithms are even smart enough to find out what you're "really" looking for and will happily detect skin color as the most significant feature for racial classifications.