In September 2016, Amazon introduced a new problem in conversational Al, the Alexa Prize, which challenged student teams to build a socialbot that could converse with Alexa on a wide range of topics. The socialbot differs from task-oriented dialog systems that address explicit user goals associated with a constrained domain, and also differs from chatbot systems that only handle social chitchat. The socialbot must handle chitchat, and needs to inform and exchange opinions with a user about recent news and other topics of interest, serve evolving user interests and implicit information sharing goals. Because of these differences, many of the conventional approaches that had been developed for dialog systems were difficult to apply to the socialbot problem. Further, because this was essentially a new problem, there was no existing conversational data that was well-matched to the types of interactions that Alexa users had with the socialbots. Initially, even applying simple machine learning was challenging, let alone end-to-end system design. Moreover, the implementation had to be scalable to a high volume of user interactions. However, access to millions of real users and the creativity of student teams led to a good start on this challenging problem.