17
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
this post was submitted on 03 Feb 2025
17 points (100.0% liked)
TechTakes
1761 readers
160 users here now
Big brain tech dude got yet another clueless take over at HackerNews etc? Here's the place to vent. Orange site, VC foolishness, all welcome.
This is not debate club. Unless it’s amusing debate.
For actually-good tech, you want our NotAwfulTech community
founded 2 years ago
MODERATORS
My sympathies.
Read somewhere that the practice of defending one's thesis was established because buying a thesis was such an established practice. Scaling that up for every single text is of course utterly impractical.
I had a recent conversation with someone who was convinced that machines learn when they regurgitate text, because "that is what humans do". My counterargument was that if regurgitation is learning then every student who crammed, regurgitated and forgot, must have learnt much more than anyone thought. I didn't get any reply, so I must assume that by reading my reply and creating a version of it in their head they immediately understood the errors of their ways.
But we know the tech behind these models right? They dont change their weights when they produce output right? You could have a discussion if updating the values is learning, but it doesnt even do that right? (Feeding the questions back into the dataset used to train them is a different mechanic)
That's true, and that's one way to approach the topic.
I generally focus on humans being more complex than the caricature we need to be reduced to in order for the argument to appear plausible. Having some humanities training comes in handy because the prompt fans very rarely do.