3 ms·
> What makes you so sure that none of the resoundingly non-crap that you have reviewed was not produced by LLM? From the post you're replying to: > Obviously
by whtsthmttrmn 2y ago
> What makes you so sure that none of the resoundingly non-crap that you have reviewed was not produced by LLM?
From the post you're replying to:
> Obviously we can't sample from unknown-authorship … nor am I; I'm sampling problems that I and others run through an LLM, and the output thereof.
- paulcole 2y agoYes, believe it or not I’m able to read. > I'm sampling problems that I and others run through an LLM This is not what’s happening unless 100% of the problems they’ve sampled (even outside of this fun little exercise) have been run through an LLM. They’re pretending like it doesn’t matter that they’re looking at untold numbers of other problems and are not aware whether those are LLM generated or not.
- whtsthmttrmn 2y agoIt sounds like that's what's happening. The LLM code that they have reviewed has been, to their standards, subpar.
- paulcole 2y ago> The LLM code that they have reviewed has been, to their standards, subpar. No. The accurate way to say this is: “The code that they have reviewed that they know came from an LLM has been, to their standards, subpar.”