It would be interesting if there's enough data to train a model capable enough to converse with and ask about contemporary views on issues of the day, or what it thought about "potential" future events/technologies yet to happen.