In my experience, getting a bot to chat about bullshit is not a particularly useful pursuit in itself. There are far more useful ML paths to explore, like guide

In my experience, getting a bot to chat about bullshit is not a particularly useful pursuit in itself. There are far more useful ML paths to explore, like guided, interactive decisionmaking in a much more focused information field. And I think these naturally will lead to authentic conversation. But to truly converse beyond an essentially overfit learned response domain requires real world knowledge(1) and learned relationships/heuristic simulation(2) in combination with decision making, and our nets aren't quite complex enough to get 1 and 2 represented sufficiently well (need lots more memory), while the decision making algorithm/architecture hasn't been developed yet... But the pieces are all availably for assembly and industry is getting pretty close. If hardware continues to scale at a similar pace we can probably expect true AI in some form in the next hundred or so years. It won't initially be very human but if it is purposed (as it undoubtedly will be) to produce improved designs, by that point progress will effectively be exponential and probably impossible to predict. All of this thanks I reluctantly admit to pioneering teams at places like Google and fb and Microsoft, and the open access nature of arxiv. I think, in proportion to its emerging complexity and value, machine learning may be one of the most quickly growing fields ever to exist. Humanity is quickly approaching a new era.