Ask HN: Have you been falsely accused of AI-generated content?

5 pointsbmaupin2mo ago7 comments

I was accused for the first time of my writing looking AI-generated. I tend to be overly verbose and use a lot of ordered and unordered lists. On top of that I'm used to documentation systems like Confluence with "info" and "note" macros so I'll often sprinkle in a limited set of emoji for emphasis when I'm writing elsewhere.

In this case the accusation was insignificant; it's for a stack exchange answer that could be deleted and my life wouldn't be negatively impacted at all. But now I can't help but wonder for the first time if the AI witch hunt is going to potentially have real consequences for some people. Has anyone experienced this in any meaningful way?

7 comments

MilnerRoute2mo ago

I wrote a blog post where I'd transcribed a conversation with a non-native speaker. Someone on Reddit was convinced the mangled syntax of the written-out words meant it must be AI-generated. (I'd also used some emdashes...)

I think the real problem is people being sure they can correctly identify AI (while they're actually just guessing wrong). Honestly, I do appreciate the efforts to weed out AI-generated content. Maybe someday we'll come up with more fool-proof detectors.

In the mean time, those weren't meaningful consequences for me. But one of the big newspapers ran an article about college students wrongfully accused of using AI, and then facing academic discipline for cheating.

Bender2mo ago

Not yet. Could it be that some people talk in the same manor as AI provides answers? Perhaps AI could have different modes of answering questions that would look more like different people, personalities, etc... Maybe this is already a thing via prompts and I am unaware of it as I am just a casual intermittent user of AI.

Me personally, I would like it if AI could give me terse answers that appear as if it were a lazy IT person that was indifferent to everything. Even better if it had a comical sidekick like the two main characters in the IT crowd. Snarky comments on its own comments.

patrakov2mo ago

I got a YouTube video in my recommendations where a professional photographer mentions exactly this problem: https://www.youtube.com/watch?v=V7v6SM8fHGQ

It is AI witch hunt now, it was Photoshop witch hunt before. Nothing new.

apothegm2mo ago

Not yet personally, but I’ve seen people who write much like I do be accused of it. And been told I write like AI/AI writes like me. (I don’t use emoji, but apparently my grammar in professional writing is impeccable, and LLMs have absorbed many of the same rhetorical techniques I was taught in school.)

rekabis2mo ago

I have a tendency to write at an above-average, technical level, so it’s already happened to me a few times on places like Reddit.

Like, it takes just a quick look at my post history to disabuse anyone of that notion, but some people are either stupid or trollish.

fragmede2mo ago

Did an LLM write this?

...sorry. It's the new world we live in and you're just going to have to get used to it. People are going to get over it eventually, or not, but "You're absolutely right!" and other tells like em—dashes are going to separate the people who care about the thing vs care about AI doing the thing. If I made you dinner, and you ate it, and found it delicious, but I used a microwave to make it, does it make it any less delicious? Now, if I'm telling you I'm a world renowned chef with Michelin stars and have an anti microwave crusade I'm on, and I'm totally a lying liar, that's one thing. But if I'm just humbly making food for a friend, and the question of microwave or not didn't come up, why does it matter if I used the microwave instead of an oven?

subdavis2mo ago

Yep, happened a few months ago. Another engineer accused me in a code review. I didn't really appreciate it, and I told them so. We talked about it, reached an understanding, and have better expectations about disclosing AI use in PRs now.

I don't think this is insignificant. IMO it's a trust and culture problem that probably needs to be addressed.

j / k navigate · click thread line to collapse