I have to keep posting this in every thread on this topic I guess. LLMs have an extremely high noise to value ratio; it is easy and quick for a user to generate

I have to keep posting this in every thread on this topic I guess. LLMs have an extremely high noise to value ratio; it is easy and quick for a user to generate an incorrect but nicely worded post and it is easy for said users to do so across multiple posts instantly. Additionally, there is no value in allowing for users to post answers by LLMs because if someone wanted that, they could've just asked the LLM themselves. A great example of this problem is people submitting bullshit PRs to repos, which has become worse with the advent of LLM popularity [1]. This means reviewers need to spend a lot more time sussing out if it's AI generated or not which takes away time from reviewing actual, valuable PRs. Or even stuff that's bad but generated by an actual human, because there's potential for actual knowledge transfer there. [1] https://mastodon.social/@danluu/110335983520055904 https://mastodon.social/@danluu/110335983520055904