A machine-vision algorithm has learned to analyze and categorize scientific figures.
In 1973, the statistician Francis Anscombe devised a fascinating demonstration showing why data should always be plotted before it is analyzed. The demonstration consisted of four data sets that had almost identical statistical properties. By this measure they are essentially the same.
But when plotted, the data sets look entirely different. Anscombe’s quartet, as it has become known, shows how good graphics allow people to analyze data in a different way, to think and talk about it on another level.
FULL STORY from MIT Technology Review