What Does Anthropic's $1.5 Billion Settlement Mean for Writers?

0:28Here's golf legend John Daly. From Taco Night in Tulum to sushi in Tokyo, make every bite rewarding with gold from Amex. Wherever you die, four times the membership rewards points at restaurants worldwide are piling up. Learn more at americanexpress.com slash explore-gold. Terms and points cap apply. All right, so today on the podcast, we're talking about Anthropik's recent$1.5 billion copyright settlement. Kind of what that means moving forward with AI and training data, how the writers feel about this. We're going to talk all about that. Before we do, Jaden, why don't you tell them about our school community?

1:01Yeah, so every single week, Jamie and I record an exclusive video that we post on our school community. We have over 60 of these tutorials and breakdowns. Actually, I've been saying 60 for like a few months. It's probably, we're probably getting close to 75. We'll let you know when we crack 100. But we have a whole bunch of different things that teach you how to do AI marketing growth hacks, Amazon influencer side hustle that we've both made tens of thousands of dollars from, how to become an AI consultant, how to do AI music, how to do podcast growth hacks, AI content hacks. There's a whole bunch of cool stuff.

1:31This week, Jamie and I recorded a whole tutorial on using Google's new Nano Banana for some really cool image side hustles that we're doing AI image stuff. And I recently sent someone a quote that I'm hoping we close for a$15 ,000 package that I did basically the whole thing using Nano Banana. And Jamie is doing a bunch of cool things with his side hustle. So anyways, if you want the inside scoop about what we're working on and the tools that we're using. Join the AI Hustle School community. It is$21 a month. We'll be raising the price next week. So if you want to lock in that price, you can lock it in now.

2:03Join this incredible community of over 300 people talking about what they're doing with AI. All right, let's get into the settlement. This just brings the biggest smile on my face. Jamie was kind of laughing before about how salty the authors of an article on TechCrunch are about it, they said, screw the money. Anthropics$1.5 billion copyright settlement sucks for writers. Anyways, they are a writer. So I understand why they feel this way. I for one am thrilled with this settlement. But I guess like just off the bat, Jamie, do you feel like like do you feel like from their perspective, it's it's fair or accurate?

2:37I mean, this is$1.5 billion. I guess what were your thoughts when you kind of saw this thing come down the line? Yeah, so the basic the concept here is that Anthropic got sued because they trained a lot of their model off of articles written by journalists. Apparently up to half a million writers are eligible for a payday of at least$3 ,000. So that to me, if I was a writer would be encouraging because, hey, I'm getting at least recognized that my work was used to train in AI and so I can get$3 ,000. bucks. That's awesome. I think the reason that some of them are salty is because they feel like, you know, they're, they've been ripped off and, and replaced by AI.

3:20So even though they're getting $3 ,000, their career's over, you know, so I could see, I could see how they kind of feel like it's a, you know, what would you say? Just like a crumb when they really deserve a whole pizza. So what are your thoughts on it? Yeah. I think that's probably their take on it too. Like my personal take is, well, I guess, first of all, who gets paid in this, I guess I'll break that down. So this is actually like, you're right. This is why the lawsuit came is because basically a whole bunch of writers were like, hey, you know, they trained off of our work and included into their model.

3:50We don't like that. The reason that Anthropic actually has to pay them isn't because they use their model to train on their work. And that's kind of what I'm happy about. The reason that they have to pay them is because what OpenAI actually did was they went and found a bunch of, They pirated a million books from shadow libraries, pirated a whole bunch of books, and they didn't even pay for them. They just went. And I think the reason why they did this is because it's really hard to and I'm not justifying it. Obviously, it's wrong for the pain of billion dollars for it. But I'm explaining why they did it is because it's really hard to get a data set of all of the all the books in the world that they want.

4:23But where the data set exists is on these pirated websites. So it's like it's hard to get it in real life. And this is evidenced by the fact that after they kind of this whole lawsuit started rolling out, the writers just wanted to get paid because they were in there. And the judge looked at the case and was like, OK, you guys aren't going to get paid for being in there. So it was Judge William Alsup. He sided with Anthropic being like, nope, they're like you're allowed to be included in their data set. But obviously it's illegal to train on material that was pirated. So he basically argued that it's legal to train, though, because the use case is transformative enough to be protected by the fair use doctrines, which is a carve out for copyright law from 1967.

5:03So basically, he's like, look, if a person wanted to go and learn about what you write about in your book and they went and read your book and then they went and wrote a paper on it because they read your book, you're allowed to do that. And that's basically what AI models are doing. This is his argument. And so, you know, AI models are allowed to do this. They're allowed to read your books. What they're not allowed to do is go and grab, you know, millions of books from a pirated website where they don't pay for them. So what's interesting is what Anthropic actually ended up having to do, which is sort of hilarious.

5:27They had to go buy pallets and pallets of millions of books, ship them somewhere where basically there's a robot that takes the book, rips the cover off of the book and photocopies every single page and just uploads the images into like a database. And then I think they just like recycle or throw away the book afterwards. so they they physically will purchase the book they physically will photocopy it and upload it and they're allowed to do that and include that into the training data but they're not allowed to go to pirated websites now you can imagine the benefits of the pirated website right it's literally text that can just be copy and pasted in it's way smaller you can immediately get it in two seconds instead of like you can imagine this whole book process was quite an ordeal but for companies making billions of dollars they're like all right if that's the loophole then that's a loophole.

6:14We'll just go buy the pallets of books, right? So anyways, this is what they eventually started to do. And then they're okay to use. So basically, they're fine to keep the content in there. They're not going to force them to like retrain their model or something. But the penalty, which by the way, there's lots of other companies, I don't know if anyone else did the shadow library, like the pirated book thing, maybe it will come out in the future. But there's a bunch of other cases, like companies getting sued, meta, Google, OpenAI, MidJourney. So I'll be curious to see kind of what what happens with those.

6:42But basically, they got a slap on the wrist. I mean, it's kind of a massive slap, in my opinion,$1.5 billion, but they got a slap on the wrist and they can keep moving forward, which I'm sort of happy about because I would hate it if like, I don't know, we lost all our AI models because some sort of thing in there wasn't allowed, kind of sucked after wait till they retrain them in X amount of time, you know? Yeah, I would be curious and maybe you can answer this question since you're more in the development world than me and stuff like that. But when it comes to training data and scraping websites and stuff like that.

7:13Do you believe that an actual human chose the pirated website and said, hey, train off this? Or was it like crawled accidentally? What does that process look like for training and data scraping? I would be really curious if Anthropik argues that it was an accident, like their bot did. I don't know if they've said anything about that. Maybe they would say that. But I think in that case, it's most definitely a human basically saying there is 1 million books at this link right here we could just download them and then they're like oh but what if we get in trouble they're like okay so there is some sketchy behavior going on it sounds like yeah that is definitely sketchy but something that's really interesting about this is that um anthropic and so i don't know if everyone else is gonna get caught up in this if everyone else did because anthropic basically from the very beginning one thing that people have always noticed and i have tons of friends ironically that are authors that love anthropic and they're like its tone is so much better.

8:08Like it understands my writing style and my tone. And I think it's like it's better tone. Basically, a lot of that comes from it had this kind of unique data set. So if you imagine like one of the main reasons a lot of people have been using Anthropic for the last few years is because the tone was better. And so if that was their only competitive advantage, they might have just been like, maybe it's worth it. So that like we basically can compete with OpenAI because it's our like only shot, right? And if they hadn't done that, their company likely would have had way less users and maybe went bankrupt.

8:34Who knows, right? Like, who knows what would have happened. So I think for them, it like it helped whether it was right or wrong, it helped their company out a lot. And I think also this I feel like a 1.5 billion dollars would have been pretty brutal. But they just announced last week that they raised over 13, like 13 and a half billion. So I think because they just raised 13 and a half billion, they can pay this, you know, one and a half billion dollar settlement, they're gonna be fine, right? It's less than 10 % of what they just raised. And they can move on. But what's interesting is right before that raise, the most that they raised was in March.

9:05They raised a$3 billion round or$3.5 billion, I think. And so, I mean, I feel like that would have been pretty brutal to have to give up $1.5 billion of your$3.5 billion round to pay like some legal thing. But now that they've raised $13 billion, it's like, yeah, they're fine. I don't think anyone's too worried about them going insolvent or anything. Yeah. I mean, I think this whole case and article just goes to show that I think moving forward, there's going to be more need for a service like Cloudflare, where they're kind of compensating creators, artists, authors based on if AI uses their data to train, at least they get compensated a little bit for it.

9:42You know, similar to the way that people freaked out when streaming services popped up and, you know, you know, how do we compensate music artists if they're being, you know, pirated all the time or streamed, you know, like, how is it going to work? I think there's going to have to be some kind of regulation or a company that comes along and is the middleman between these two these two groups of people. So yeah, it'll be interesting to see. Yeah, 100%. Hey, if you guys learned anything new on the show today, we'd really appreciate it if you could leave us a rating and review on the podcast, or if you just appreciated it helps the show out a lot just get shown to new people and for us to grow our audience to more amazing people like yourself.

10:22And also make sure to go and check out the Aihustle school community. We'll leave a link in the description. If you've ever wanted to join it, this would be of the week. We're raising the price next week. And so you could lock in the price now. Right now, I think it's$21. We'll move it to 22 and it'll keep going up every week. So anyways, if you want to lock in the price, do it now. We'd love to have you as a member of the community and share everything that we are working on with you. Thanks so much. And we'll catch you in the next episode.

10:51Welcome to Moto Casino, where the excitement never ends with thousands of the hottest free Free-to-play social casino games, fastest payouts, and the best promotions in the industry. No tricks or gimmicks. Owned and operated in the USA. Moto Casino is a free-to-play social casino. No purchase necessary. 21 plus to play. Void where prohibited. Sign up today for a generous welcome bonus. Moto Casino. American Social Casino. Download the Moto Casino app today. From Taco Night in Tulum to Sushi in Tokyo, make every bite rewarding with gold from Amex. Wherever you dine, four times the membership rewards points at restaurants worldwide are piling up.

11:27Learn more at AmericanExpress.com slash Explore-Gold. Terms and points cap apply.