← Home ← Back to /pol/

Thread 512581870

8 posts 4 images 7 unique posters /pol/
Anonymous (ID: JHC6AbcC) United States No.512581870 >>512582011 >>512584095 >>512585690
AI is a glorified reddit user
Anonymous (ID: M+WI4kPe) United States No.512582011 >>512584477
>>512581870 (OP)
I read some dude trained an LLM on a years (?) worth of content from 4chan and it turned out exactly like you’d expect. Pretty much every other LLM engineer condemned him kek. I wish I could use it.
Anonymous (ID: bn+Q3bLm) United States No.512582974
>bots teaching bots
>Billions being poured into AI
>Never learns from a real person
AI will come and go, useless every step of the way outside of novelty
Anonymous (ID: ytEqvsm+) United States No.512584095 >>512584356 >>512584362
>>512581870 (OP)
Is that AI generated? Those numbers don't add up to 100%
Anonymous (ID: YUQQLu0g) Serbia No.512584356
>>512584095
one keyword can have multiple citations
so 40% cited reddit, 26% cited wikipedia, but not exclusively
Anonymous (ID: M+WI4kPe) United States No.512584362
>>512584095
I’m thinking because it’ll cite multiple sources in one message sometimes.
Anonymous (ID: BVrmhGsM) United States No.512584477
>>512582011
I ran llama 3 with it on my laptop
And it'd slowly grunt out..

KILL
ALL
NIGGERS

Or whatever. Pretty good
Anonymous (ID: X8h/DGZW) Canada No.512585690
>>512581870 (OP)
What this means is that you can hack AI consensus by socially engineering your posts to the top of its citations and tailoring them to be directly relevant to specific queries. In fact, there are already groups doing this as an extension of the Wikipedia wars that defined public opinion through the 2010s.