A plausible answer tells you what someone might say. How many people would say it is a rather different question. Introducing Snowglobe: population predictions, synthetic respondents and a way to follow the same inquiry from a national estimate to a conversation or a map.

An embedding puts something into a vector. What we can reasonably expect to do with it afterwards is a more interesting question. In building Giedi, we wanted to represent populations, perform operations on them, and use the results to answer questions we had yet to ask. This is an account of how far we got.

The year is 1990. A beastly wall casts a great big shadow on the once-great, now-convalescing Holy Roman Empire. Gorbachev and Bush Sr., presumably under the influence of unholy amounts of libations, set about reordering the geopolitical landscape with all the finesse of bulls in china shops. Most concerningly, vexation plagues the connectionist community. Enter stage left one Jeffrey Elman, Finding Structure in Time clutched firmly in hand.

Tuesday, September 11, 2001, dawned clear and pleasant, as the red disk of the sun broke the horizon where the Atlantic Ocean met the sky. As Command Duty Officer (CDO) on the aircraft carrier USS John F. Kennedy (CV-67), I was up early to meet the day. Though I was not expecting it, this day turned out to be one I would never forget.