Interpolation

Is that what it does? Yeah, see? You couldn't figure out the extrapolation equation either. Did I write that one down or no? There has to be a way convey that when it is engaging in conjecture that it's running it's training data against a finite/infinite of potential data... Hym "I mean, I guess that's the difference between interpolation and extrapolation. Interpolation is N and extrapolation is N(->Interacts with somehow->)U = Output (at whatever probability). But that is why I told you to do that thing from that movie or game with the robot that qualifies it's statements before speaking. Like 'Conjecture It could be that something-something whatever' or 'Witty retort That's what your mother said last night.' Those hallucinations would be getting a Conjecture label if you had done that and then, at the very least, the user would know which outputs are less likely to be true. That isn't unlike a human mind though. People be making shit up all the time. Seems like a problem that could be solve with a 'could be X but IDK. Need more data.'And then I would be like 'What data would you need to draw a more accurate conclusion?' But, whatever. Security?