Bob Novella

So that would be good as well.

Okay, so let's go to now today, most recently, Thomas Dietrich.

He's the chair of the archives computer science section.

He posted just last week.

He said,

So the key here is obviously incontrovertible evidence.

So what do you think he's referring to when he says if they find incontrovertible evidence that the authors did not check the results of any LLM generation?

What do you think that incontrovertible evidence is?

There's mainly two classes of this type of evidence that he's referring to.

Yeah, so they read the paper and they find evidence, solid, I mean, absolutely solid evidence that they clearly did not check their LLM assistance in this paper.

Yes, that's the biggest one.

The hallucinated references.

It's so common.

It's so common.

There's been so many retractions in papers, mainly because of these AI hallucinated references.

And so that's a huge one right there.

And when I say incontrovertible, that's clearly incontrovertible.

Because if you read a paper and there's a reference, and that reference does not exist and never has existed, then clearly this is something from an LLM.

you know i mean with overwhelming certainty right there's there's one that's even better even more incontrovertible than that and that one is that you actually read the paper and find comments to or from the llm so one example here is that if you're reading and you find something that says this here is a 200 word summary would you like me to make any changes i mean if you read that

Clearly, they totally screwed up.