søren peter mørch

Conversation

Recent posts in reply to #22qa4ca

bender@twtxt.net
@prologic that's not the way LLM work. LOL. If they were to "learn" by user's corrections, or contributions, then they would really be screwed.
In reply to: #22qa4ca 3 months ago
movq@www.uninformativ.de
@prologic Sometimes some “coding AI” spits out some partially useful snippet, and then one of us nerds panics and claims the end is nigh. I think that’s about it. 🤔

All these “AI” topics are extremely boring to me. (Traditional) computer stuff is a hobby and a passion of mine because I can *understand* it. How does an operating system work, how does a web server work, how does a keyboard work, how do you do 3D graphics, … All those are things that I can learn and understand, and that’s the thrill. But anything related to “AI”? It’s just statistics and a large model that spits out something for some reason … Super boring. I couldn’t be less interested.

And on top of that, “AI” requires a ton of energy (as far as I know), which we, as a species, cannot afford right now. It needs to stop.
In reply to: #22qa4ca 3 months ago
bender@twtxt.net In reply to: #22qa4ca 3 months ago
bender@twtxt.net
Only Llama 2 of my locally running LLM got it right:
!Llama 2 correct answer
In reply to: #22qa4ca 3 months ago
Comment via email