About a year and a half ago, I wrote about my kid’s experience with an AI checker tool that was pre-installed on a school-issued Chromebook. The assignment had been to write an essay about Ku…
i’m kidding in this case; i actually did read your comment.
but, more and more frequently i do find myself stopping reading something due to suspecting (broadly speaking, because of the “quality of the content”) that it is likely to be LLM output.
I find I’ve been doing this long before LLMs though. A lot of time you open an article, and the writing is just terrible, you try get through a few paragraphs and just give up. Most content online just isn’t well written or that interesting. You’re right that text that’s clearly LLM generated tends to be a good indication of low quality though. But I find the tells come not from the writing style like using em dashes, but from the superficial quality of the content. LLMs are not good at generating any depth or actual analysis, so the text that’s LLM generated reads like a bunch of sound bites strung together.
ai;dr
i’m kidding in this case; i actually did read your comment.
but, more and more frequently i do find myself stopping reading something due to suspecting (broadly speaking, because of the “quality of the content”) that it is likely to be LLM output.
I find I’ve been doing this long before LLMs though. A lot of time you open an article, and the writing is just terrible, you try get through a few paragraphs and just give up. Most content online just isn’t well written or that interesting. You’re right that text that’s clearly LLM generated tends to be a good indication of low quality though. But I find the tells come not from the writing style like using em dashes, but from the superficial quality of the content. LLMs are not good at generating any depth or actual analysis, so the text that’s LLM generated reads like a bunch of sound bites strung together.