this post was submitted on 04 Jun 2023
27 points (100.0% liked)
Technology
37654 readers
382 users here now
A nice place to discuss rumors, happenings, innovations, and challenges in the technology sphere. We also welcome discussions on the intersections of technology and society. If it’s technological news or discussion of technology, it probably belongs here.
Remember the overriding ethos on Beehaw: Be(e) Nice. Each user you encounter here is a person, and should be treated with kindness (even if they’re wrong, or use a Linux distro you don’t like). Personal attacks will not be tolerated.
Subcommunities on Beehaw:
This community's icon was made by Aaron Schneider, under the CC-BY-NC-SA 4.0 license.
founded 2 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
Well, for language model, it basically regurgitate whatever it remembered from training data with some noises, so some information might be correct from the training data, but when it is generating something that it wasn't trained on before, then it could present incorrect answer. I only really use ChatGPT for generating documentation, to make it sounds better and flow easier for the readers.
Not could, will. It's basically guaranteed to start spitting out garbage once it's extrapolating beyond the training data. Any semblance of correctness is just luck at that point.
This is true for basically all models, everywhere.
Not really, because there were some reports of world model being utilized alongside with LLM. While it's early and premature, it does indicates that we're going to see world model being incorporated with the language model so in due time, it could conceptualize the world.
"Well, if we just include other data", weirdly, isn't a rebuke to "it can't extrapolate beyond the data".