Teaching Robot Dogs Linguistic Tricks
While written from the perspective of engineering, this short article mentions several important features of human language, including the “conventional” (or “arbitrary”) nature of linguistic signs. On the face of it, the outcome seems rather limited in that these robots create lexical items instead of linguistic structures, but the very idea that robots can learn from one another is a fascinating one (and a well-known concept for artificial intelligence).
Also, it’s nice to see that the EPFL is involved. With its new supercomputing power (one of the fastest academic computers in the world), l’EPF has other reasons to be appreciated.
Disparate 
Thanks for the precisions and link. That Sony site seems to be down right now but it should be worth a read if there is more to the project than “start[ing] from scratch to develop common agreement on a word to use to refer the ball” (as per the short Engineer Online piece).
What do you mean “only lexical structures”? If you read the papers, you’d see that they create grammars as well. See Luc Steels work at Sony’s CSL laboratory http://www.csl.sony.fr/