Well, I can see how that could happen, and in fact, copy-paste artifacts and unintended summaries/hallucinations have happened to me when grabbing output back from an LLM.

Here's the thing though: I catch it 100% of the time because my writing has version control and I compare diffs. When dealing with something that can exist as plain text, there isn't a good reason not to have that setup. I'm no journalist, but it blows my mind that writers who deal specifically in reported facts apparently don't have systems in place to idiot-proof and preserve their sources of truth.

I get it, at some point back in the analog days there were more editors and copywriters that actually verified these things, and those jobs were sacrificed at the altar of capitalism. I've seen writing quality on the web take a downturn as a result. But for fuck's sake y'all, maybe do the bare minimum and start implementing safeguards before you let your writers use inherently lossy tools?

Thank fuck Bsky has these character limits or else he would have had to put all that text in an easily legible format for reading and copying. Fuck character limits up their stupid asses.

This sounds eerily familiar…

I don’t know if Hearst told him to use a chatbot to generate their “Best of Summer Lists,” but it doesn’t matter. When you give a freelancer an assignment to turn around ten summer lists on a short timescale, everyone understands that his job isn’t to write those lists, it’s to supervise a chatbot.

But his job wasn’t even to supervise the chatbot adequately (single-handedly fact-checking 10 lists of 15 items is a long, labor-intensive pro­cess). Rather, it was to take the blame for the factual inaccuracies in those lists. He was, in the phrasing of Dan Davies, “an accountability sink” (or as Madeleine Clare Elish puts it, a “moral crumple zone”).

https://locusmag.com/feature/commentary-cory-doctorow-reverse-centaurs/

This is a good way to handle the situation and an understandable and believable scenario, so I'm perfectly willing to forgive this. I'm a little less okay with an apparent "work in spite of illness" policy, however.

But still, it's a serious blunder, and it needs to be said that any repeat of this at all would be very damning. I can't forgive this level of fuckup twice. Any AI use is a risk, folks; treat it like one.

Why would he play with an AI toy while he's doing his job and he's sick?

Of course something was bound to happen.

Thanks for sharing, I was wondering if he would say anything about it. They seem to be handling it well.

We are reinforcing our editorial standards following this incident.

It sounds like they will be reminding their team not to do that and scrutinizing articles in the near future

The author explained

Such a dangerous game for them to play. Ars is its reputation. If it's just another slophouse, what's special about it? There are millions of others. Or I can go to the source and have ChatGPT generate the slop.

I think there’s still reasonable debate over whether or not a human actively triggered the agent to generate the hit piece.

This whole story is a little bit confusing since there are two entirely separate ethically questionable uses of AI by different parties, the person running the rogue coding agent, and the author of the Ars article using AI to misquote the victim of that coding agent.

Or… he was ill and made a stupid mistake as he should have been resting?

Either it was a once-off or Ars will fuck up again. Guess we’ll see.

Benj Edwards and Kyle Orland were credited as the authors