Post #3261878

Let's turn to the second plausible use case: customer support. People often like to talk to other people to resolve problems. What if we can automate the *feeling* of talking to a person, but with no actual people involved? There are a bunch of things going on here. At least here, there's a real user need. But to what extent is it a real user solution? It's plausible to me here that this will be a plausible first-query solution once you've built up a good base of Q&A examples for it, doing a bit of textual generalization. At least as long as what you're doing is pretty standard. But remember that we're using a bullshit engine here, so what happens when the statistically plausible text isn't correct or useful? 6/