In late 2014 Google announced a deep learning system that could automatically generate a descriptive caption of an arbitrary photograph.

A significantly greater challenge is to provide an understandable and comparably accurate description of the events taking place in a short video (without audio included).

Will a working version of such a system be publicly announced and demonstrated by July 1, 2016? For success, the system should act on a diverse set of short videos similar to these, and should be comparable in accuracy to the accuracy of Google's image captioning when it was first announced.