Technology

5% of the time? LLMs, from their own perspective, are only capable of hallucinating. There’s no difference in what they’re doing between cases we call “hallucinating” and “correct.” It’s the same process.

I think the executive in question is Kyle Orland, who I don't know personally but I've interacted with sometimes. He's pretty good! Again, as I've said elsewhere in this thread, maybe I'm too close. I've never worked for either of them, but I've encountered them on social media from time to time. I think I interacted with Kyle concerning a Storybundle book once.

Copy editing won't be an executive's job. But yeah, they didn't do the bare minimum which is concerning, it seems to indicate that they may not do the bare minimum on all of their articles. How much stuff went undiscovered?

I'm not going to outright say that journalist shouldn't use AI to write articles, because it's basically an enforceable rule, but there should be someone at some point whose ultimate responsibility is to make sure that the articles are at least factual, whether they were written by a human or not. Determining whether a quote is legitimate is pretty easy, you just have to Google the quote, if you can't find any other sources you start to ask questions. As I said it's the bare minimum they could have done.

The executives will get a bonus this year.

well of course! they just saved a lot of money on wages, they deserve it!

The comments here around this were so... Off. I guess nothing was certain, but we were supposed to believe that the author was too sick to write an article, but also writing an article and using an AI "tool" at the same time.

Hindsight is 20/20, but popular defenses at the time were

He wrote the article himself, he just got mixed up when experimenting with using an AI tool to help him extract quotes from a blog entry. (He is the head AI writer, so learning about these tools is his job.) It was nonetheless his failure to check the quotes he was copying from his note to make sure that he got them right… but an important bit of context is that he had COVID while doing all this.

I did not downvote you—my instance does not allow or show downvotes, which is really nice!—but he was sick, and he did make a mistake, and him being fired does not make either of those things false.

Also, a ton of people were piling on him in that thread, so you had plenty of company in calling him to be fired.

It seems like he had humility, but he put his name on an article that had false content that he didn't verify. That's not a mistake so much as it is neglect of due diligence. Simply checking if the important citations in his article were true would have saved him, but he didn't. I can only imagine how many journalists do this without getting caught.

Absolutely not. Ars has a no AI policy, it's the exact opposite. Guessing you are a nice little bot.

To be fair to Ars Technica, that doesn't sound like the case to me.

The "journalist" in question seems to be suggesting that this was their own bad judgment to use AI to "find relevant quotes" from the source material.

Having said that, there's also a senior editor on the by-line who hasn't been held accountable for clearly failing to do their job, which as I understand it, is to read, edit and verify the contents of the article. So in a way Ars seems to have a problem with quality whether or not the use of AI was mandated.

Maybe they should do more than just fire a person who was caught using AI. Maybe they should establish a process of independent fact checking before publication, regardless of whether AI was known or intended to be used to produce the article. It is a problem that AI was used in a way that introduced factual errors. It's fair that the person responsible for this was fired. But all processes need quality control. Why hasn't the person who failed to wrap quality control processes around the author fired?