[II2II]

I don't downvote those comments, even though I have a serious problem them.

These comments are little more than a witch hunt. It is clear from the language being used: "it's obvious", rather than providing evidence. When people do provide evidence, it is in terms of "tells". In other words, bits of overused grammar that are common in LLM generated texts yet also exist in human written texts. It doesn't really prove their claims. Worse yet, it is also next to impossible to defend one's self from such claims.

None of this means that I want to spend my days reading LLM generated articles. I believe they pollute the Internet with yet another source of hollow writing. (LLMs are not the only guilty party here. Plenty of flesh and blood humans do the same.) They also further necessitate the use of LLMs for what I think is the one legitimate use, which is research. (Before you attack LLMs for hallucinating, it is worth noting that many, if not most, of the articles written by people demonstrate the same.) Finally, if I am interested in the output of an LLM, I would rather do it myself. At least then I would know what I am getting is LLM generated, rather than a misrepresentation. Plus it is easier to dig deeper if something seems to be out of kilter, either through further prompting or requesting sources.

Yet all of my distaste for LLM generated articles does not outweigh my distaste for the witch hunt.