Karen Hao

I have so many questions after seeing this video, because the first thing that I want to know, I mean, Farlands is such a brilliant name. I want to know who even came up with that term. But also, what are the types of videos that are populating the Farlands, and how do they get there? Are people producing content specifically to land in the Farlands, or are they just producing bad content, and then it gets damned to the Farlands, but now there's an entire community regrouping and reclaiming the Farlands? What's actually going on?

Luulen, että Farlands oli suunniteltu, jolloin ei ollut paljon ihmisiä, jotka pystyivät käyttämään sitä, mutta onko se oikeastaan suunniteltu vain non-mainstreamin suunnitelman? Onko se siksi, että se ei ole enää Farlands? Ihmiset ovat yrittäneet sen ja se on...

that only a private group of people appreciate. The aesthetic of these videos reminds me of this other thing that's been happening with generated art, where there's this kind of creepy, goalish woman that will sometimes appear when people have spent too much time. What they do is they'll generate the art, and then they'll pop it back into an AI generator and be like, make it more...

This or make it less that or whatever. And they'll just do it in this crazy cycle all the way down the rabbit hole until suddenly this like particular artifact appears. This freaky messed up face of a woman. I mean, that could absolutely be a Farlands. Yes. So this is Loa. And yeah, like she has a name. And so there's this whole genre of

Mythology, I guess, that has spun out of this artifact. But the moment I saw the Farland Studios, which I've literally never seen before, I was like, oh, it reminds me exactly of Loa. You're listening to the documentary from the BBC World Service.

Jep, puhutaan AI ja vihreää asiaa. Olemme saaneet todella mielenkiintoista kommenttia kuulostajalle. Tämä on Gracein, jossa hän puhui siitä, miten hän toimii ekonomisesti vahvistuneilla opiskelijoilla maailmassa, jotka tuottaa heitä kansainväliseen koulutukseen.

She started noticing this phenomenon where students would write an essay for an application to an institution that was wholly their own. No AI assistance, no AI generation whatsoever.

Ja sitten, koska he olivat miettineet, että heidän omaa kirjojaan saattaisi jotenkin saada AI-kysymyksiä mahdolliset automaattiset tutkimustietoja, he painoivat niitä näihin AI-tutkimustietoihin itseään, jotta he pystyisivät muuttamaan niiden kieltä, jotta se näyttäisi vähemmän AI-kysymyksiä. Mutta sen jälkeen...

Changes were actually really bizarre. It wouldn't actually make the essay sound better or make it sound more human. It would just evade the AI detection software and introduce all of these new artifacts that were just actually not good writing.

And she wrote to us saying, you know, like she's tried to explain to these students, you don't have to do this, it's not necessary, but they just can't get over this fear that because AI detection software is being used more and more and all around us, and they have a higher chance of getting flagged as English as a second language speakers,

They just insist on going through this process and actually degrading and devaluing their own writing even more. They're making their own writing worse? That's so depressing. Yeah, so this gets to a really interesting question, which has been coming up a lot recently, because...

There's this new AI detection software called Pangram that has become really, really popular. And recently there's been a spate of headlines about people being accused of having their writing being partly AI-generated or even all of it AI-generated because of people pasting that writing into Pangram.

Ja Pan Graham sitten arvioi sen ja tarkoitti, että kyllä, siellä on jotain AI-generointia. Joten puhuttiin pohjasta viikossa ja hänen ensykliikkiänsä, jota rakastin. Siinä oli arvioituksia, että jotkut hänen kontenteistaan, jotkut hänen paragraafistaan, olivat 40% ja 100% AI-generoituja Pan Grahamin vuoksi.

There was also a particularly famous example. Have you guys heard of the novel Shy Girl? No. So this is a horror novel that was supposed to come out and got pulled. Oh, yeah. Because of accusations. Yeah, because of accusations that it was...

AI generated and the pangram executive team also weighed into this discourse and they were like, yes, our tool says in fact that a significant amount of it is AI generated. And was it? Did it turn out that it was or was this a...

This is the problem. The writer, of course, absolutely denies it. She did mention at one point that she believes someone that she asked to help edit and review may have used some AI tools in the editing and reviewing process, but it was not her doing that. This is the challenge, is that

You can't really know. After that, there was a kind of a little wave of people putting classic literature into these AI detectors. And I think it found that like the first half of Moby Dick is AI generated. So yeah, so there's like all of these examples, but a lot of them end up pointing back specifically to pangram because pangram claims that

that it is the most accurate AI detection tool on the market and is becoming more and more popularly used, especially in academic contexts. And so the listener that wrote to us, Grace, her students were actually onto something. They are increasingly moving through a world that is being used

controlled by this kind of software. And what's interesting about this software is that even for Pangram, which there was one independent study

Not yet peer-reviewed, but that did find that Pangram generally has a pretty high accuracy rate. Pangram says itself that it's false positive rate, so falsely identifying human-generated text as AI-generated text is 1 in 10,000.