in there and say panogram but it's just pangram I think it's just pangram yeah okay so tell us a little bit about pangram and then I want to provide people with an action that they can take as well

Yeah, absolutely.

So Pangram is a company that uses LLMs, uses AI to create an AI detection software.

So everyone's probably at least a little familiar with AI detection softwares at this point.

You paste in some writing and then it tells you if it's AI or if it's human written.

And a lot of them that are out there are complete crap.

So whatever you put in, you can put in like Mary Shelley's Frankenstein and they will say, oh, this is AI written, but for a fee, we can humanize it for you.

Where are you with your use of AI with your writing or as a creative?

When I get to have the pleasure of sitting down with somebody like you, I always just want to pick your brain about what, And how do you create that imaginative storyline? Like, where does that come from? What is your source of inspiration? Do you feel like you were born with it? Is this a skill you developed over the years? Is it a combination of both? Like, can you walk me through that a little bit?

Yeah, I love that question.

And I want to hear Rosie's answer, too, because I know you have a completed draft manuscript at this point that I have not read yet, but that I have heard the buzz is that it's fantastic.

But for me, I grew up as a reader like you.

I would go to the library every week and I'd come home with a huge tote bag full of books and swap it out every week.

And as a little kid, I started just trying to emulate them.

So I would sort of, you know, write and illustrate my own books and mimic and ape like the kinds of stories that I was reading and how they were written.

Yeah.

Yeah

... now AI in this arms race that can make it look like somebody is writing.

So instead of just shooting out text, uh, the way that like a, you know, ChatGPT would, it does it to em- emulate human writing and it go, it ba- does backspace and it introduces errors and it moves things around.

So even a sort of tracker won't necessarily be able to catch, the way that somebody, the way that a ChatGPT would generate text versus the way that a human would generate text.

It's terrifying that this is what we're, like, giving up our fresh water for, this incredibly stupid arms race.

But I don't think there is a clear way to prove it, unless you're writing everything by hand.

You can certainly prove a positive if somebody has said they use AI or if somebody has, for example, left AI prompts in there.

Andy Martz, what is your rant?

I'd love to hear your response to what we heard there from Coral.

Um, I have qui- [chuckles] quite a few notes jotted down.

Um, a few things that I wanna focus on, though.

I'll start with the Anthropic lawsuit because, uh, obviously I'm very involved with that.

So Coral is correct that the judge in this case decided or declared that if Anthropic had obtained these books legally, it would've been okay for him to... for them to, um, feed them into the algorithms to create these, uh, LLMs that we're all using.

But the fact is that across all of the major, um, you know, LLMs that we're using, they did use pirated books.

We cannot ignore that.

Um, w-what do you think about them and, and just the idea that in order to detect AI, we're using AI?

Andrea, w- for people don't know the story, um, what happened with, with Shy Girl? Why did this suddenly explode, uh, into a thing?

So my understanding is that this book was originally self-published, and it got a decent amount of attention, a lot of readers.

And as is fairly common, publishers kind of took notice, and Hachette specifically picked it up to publish it in the US this, this month actually, uh, in, in April.

And a UK publisher also, um, put it out in e-book and in, uh, print in the UK.

So it had already been sort of circulating in the world when an individual reading it, uh, noticed that there were some accusations of it being written by AI.

There were people making TikToks, people claiming that sort of the rhythms and patterns of the writing sounded like generative AI.

So this individual, um, took the manuscript, ran it through several AI plagiarism checkers, uh, and brought his findings that this appeared to be AI written to The New York Times.

Mm

So I wanna, uh, uh, I wanna get your thoughts on whether more categories or certain categories of writing are more susceptible, and if you feel it's in a way unfair competition.

Yeah, absolutely.

So I will say, as the models get better and better, I do think that, uh, no author will be safe.

I think even a nonfiction book, um, could be potentially AI-written.

I mean, think about a student writing a, a paper for school.

Uh, AI...

A, a nonfiction book might not be that different [chuckles] once these models get good enough to really sound like an author putting together an entire book.

How do you feel about maybe releasing your books with a label, um, you know, not written with AI?

So I wanna, uh, uh, I wanna get your thoughts on whether more categories or certain categories of writing are more susceptible, and if you feel it's in a way unfair competition.

Yeah, absolutely.

So I will say, as the models get better and better, I do think that, uh, no author will be safe.

I think even a nonfiction book, um, could be potentially AI written.

I mean, think about a student writing a, a paper for school, uh, of AI.

A, a nonfiction book might not be that different [chuckles] once these models get good enough to really sound like an author putting together an entire book.

But I agree that right now, it's these commercial genres, like you said, like mysteries, like romance, that do seem a little bit more susceptible, and that makes sense to me.

How do you feel about maybe releasing your books with a label, um, you know, "Not written with AI"?

this company had just gone to just a book firm, bought a download and taken it from there.

So that's a really interesting question because the judge in this case, and of course there are dozens of different copyright lawsuits against AI companies that are working their way through the courts now, and so this is just one.

In the case of this lawsuit, the judge declared that if they had obtained this stuff legally and used that to train their algorithm, that would be fine.

And indeed, what Anthropic did is, which I think shows sort of their culpability, if anything, is that after they had spent some time learning from this algorithm they'd created and this large language model, they said, this feels a little sketchy that we did this on pirated books.

This might get us in trouble down the line.

And so they paid... tons of money for people to go out and buy hundreds of thousands of books in mostly used form.

So again, the authors didn't see a dollar, but they picked up used books.

Andrea, how did you acquire how many hundred thousand fellow litigants here?