It's common to assume that the ideal way of communication with robots will be verbally but this isn't always the case. Sometimes it's easier to teach by showing examples. A group of Tokyo researchers from the JST ERATO User Interface for Design Project demonstrated this technique at Siggraph 2009. The YouTube video above captures the final results, showing the robot successfully folding a garment. A few more details of the graphical training process can be found in the one page poster for their exhibit (PDF format).