Interpolate

From Latin interpolare, Latin interpolatum, from inter (“between”) and polire (“to polish”).

Interpolation provides a means of estimating intermediate points.
In other words, it's estimating in-between values.

Let's interpolate to get in-between the values.

by millennIumAMbiguity January 24, 2022

Flag

Get the Interpolatemug.

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?

I have no idea what would stop prompt injection but the admin status should stop that from working in a away that is detrimental to ME at least.... But that doesn't help YOU... Abstraction... The hierarchical model might not have as much of a problem with it (Also my idea and/or the Chinese are better at following instructions). I watched this guy train an AI to play a racing game. And to train the AI he uses Naruto shadow clone jutsu training except the growth was linear rather than exponential and it relied heavily on random occurrences and an artificial incentive structure to inch closer to the goal. So it dumps out a bunch of clones, attempts the goal, and then one of the clones will randomly and accidentally do something that brings it closer to the goal, and then that clone with be the baseline model for the next batch of clones. But after watching it I wasn't convinced that that was the best way to do that because it doesn't really have a process by which it figures out how to achieve the goal but rather just kind of stumbles closer and closer to the goal until it gets there... Might not be the best way to do that. There are still a lot of questions there for me. Faster I guess for the user but the AI doesn't actually learn anything. It's more like 'find a lightswitch in a dark room' But if you wanted me to do this you should have done it BEFORE an infection ate my brain for years. You should have interpolation...Ed better. You could try abstraction jigsaw puzzles."

by Hym Iam October 19, 2025

Flag