Brains on the Outside: A Business Ideas Podcast

Lauren Beukes: How to Destroy AI

• Alex and Andrew • Season 2 • Episode 8

Share episode

Lauren Beukes is the author of award-winning novels like Afterland and Zoo City, with her book Shining Girls was adapted to an Apple TV+ show. She has asked us about how to poison AI and whether there are methods to get it to stop stealing everyone's work. 

Some of our potential ideas: Fill books full of ads, destroy technology and go back to town criers, or put every book in your house into one big book. 

Thank you Lauren for getting in touch with us. How would you solve their problem? Email us at: brainsontheoutside@gmail.com 

Instagram: https://www.instagram.com/brainsontheoutside/ 

Tik Tok: https://www.tiktok.com/@brainsontheoutside 

Thank you to Rich Endersby-Marsh for our theme music: https://soundcloud.com/rich-marsh 

Keep your brains on the outside! xo 

Alex:

Hey, Andrew.

Andrew:

Hey, Alex.

Alex:

What's this?

Andrew:

This is brains on the outside.

Alex:

What's that?

Andrew:

Well, brains on the outside is a podcast where we dream up ridiculous solutions to the world's biggest business issues.

Alex:

Oh, yeah, I remember. Each week we take an impossible problem from a real business, and we just solve it.

Andrew:

This week, our problem is from, Lauren Beukes.

Alex:

Wait, Lauren Beukes, the author?

Andrew:

Yeah, the one with the show on Apple TV, the shining girls.

Alex:

Whoa. How much we getting paid?

Andrew:

Well, as, ever, Alex, we are doing this out, of the goodness of our hearts. There's no sponsorship bribes or perks. Just unfiltered, amazing business advice.

Alex:

So nothing.

Andrew:

Nothing. Across the season, Alex, we've helped a lot of real businesses, right? Yeah, we've helped all birds, with their shoes. We've had phone case manufacturer wave make their phone cases more unique, more sustainable.

Alex:

Yeah.

Andrew:

But this week, it's. This is. This is a bit of a twist.

Alex:

Yeah, this is, I feel this is a, Allbirds is a real big one. That's a big company. I mean, every company that we've worked with has obviously been amazing and wonderful, but this week, we have someone who actually has a tv show on Apple TV has asked us a question.

Andrew:

It's pretty exciting.

Alex:

It's really exciting. So Lauren Beukes. And that is how you pronounce her name? Because she did send me a message that said it's bucus. Like mucus. So that's what we're going with. she's an author. She's got the shining girls on tv, and she has asked us about AI and specifically, I guess, about what to do about it. She says, as a writer, how can I help? Poison the big tech LLM, plagiarism engines, aka AI, that have already been built on my work. Six of my publications are in the books. Three pirate database, which is used to train them. That's a good, interesting question.

Andrew:

Thank you, Lauren, so much for asking that question.

Alex:

Yeah, thank you so much.

Andrew:

We see a lot of AI in our jobs, in real life, behind the curtain, you know, we're pretty AI cynical.

Alex:

I would describe it.

Andrew:

yes. Yeah, I like how Lauren as well. She's jumped right to. There's no compromise. There's no discussion. We're gonna go and mess that machine up. Yeah, I love that vibe as well.

Alex:

That's like, jump to the end. Yeah, no negotiating. We're going straight in. All in. Let's screw it up. I really like that. I can see how it would be extremely annoying to have books that you are writing being used to automatically write books that computers are writing that are not as good.

Andrew:

Have you ever checked to see if an AI model can replicate this show?

Alex:

No. That would be quite exciting, though.

Andrew:

Well, I guess we were working hard to get transcripts up.

Alex:

Yeah.

Andrew:

But potentially we're not in the system yet because only a few episodes have transcripts. I assume they have been scraped though.

Alex:

Holy moly. I've never checked to see whether chatgpt knows who we are.

Andrew:

I have actually no idea. I live in fear, a little bit of it. I know that sometimes at work people point out that there's enough audio out there of us that they could effectively make us say anything they wanted. And, I know we all laugh and have a good joke about it, but Abamey's like, well, please don't. Please don't do that.

Alex:

It's not a bit of you, though, like, well, you could probably make us say anything you wanted in real life anyway.

Andrew:

Yeah.

Alex:

Just ask us nicely enough, that would be okay. You don't need a computer to do that.

Andrew:

No, we're. Even though we say we don't take bribes up front on the show, we probably would if it was good enough.

Alex:

Exactly. We did that episode with the auto trader ad in it.

Andrew:

Oh, we have taken a bribe. That's right. Oh, my gosh.

Alex:

Yeah. Yeah. I'm not sure they got what they wanted, though. I'm really honest.

Andrew:

No.

Alex:

Yeah. So, this AI problem, what do you think about it?

Andrew:

I think it's a very interesting space. I think that we have some kind of some bounds already for her exploring this idea. She wants it poisoned. Yeah, she wants it taken down somehow. Yeah, that's quite an interesting space. Might be useful to broaden it a little bit to begin with, go a bit wider, because there's a lot of interesting stuff about how to, well, one poison it. But a really interesting space is how do you make the podcast? You made the tv show, you write the book, you write the song. You saying just incomprehensible to the machine so that even if some Silicon Valley weirdo tried to take it, they couldn't. So I think there's some interesting spaces we can explore here.

Alex:

I agree. I mean, there's that. The thing that immediately came to me, the first idea was you just hide a bunch of invisible words in your book. I remember way back when there was a story about somebody writing in the bottoms of their emails in white font the word spoons a thousand times and sticking that in the bottom of their email that they sent to a friend. So every time their friend opened their emails, all of their ads, their Google Ads were all about spoons the whole time.

Andrew:

That's amazing. That is so malicious.

Alex:

So there you go. That those blank pages at the beginnings and ends of books that they leave in there just fill them full of blank words.

Andrew:

So in a lot of documents you get, there's usually a page. And sometimes in books I have afterland, one of the books right next to me. Let's. Hold on. Let's have a little look. I count 1233 blank pages before you get to chapter one. That could just be hundreds of words.

Alex:

Just Zuckerberg smells. Just over and over again, over again, just over again, over again. Every book starts with Zuckerberg smells, and then it starts the book you've also chosen.

Andrew:

I guess Zuckerberg does have his own large language model, but he didn't go for the open AI guy.

Alex:

All right. Yeah.

Andrew:

And we as well. If you're on a Kindle, you have the blank spaces, right?

Alex:

Yeah.

Andrew:

You know, you could just type in white in all those spaces as well.

Alex:

You totally could. In between all the words. Every word has between. It has a short, small, very small font, invisible word. Also on a Kindle, you can tell it when the end of the book is. And then you could just have 2000 pages of crap after that. Oh, yeah, you don't, like, you can reach the end and it's like, well, actually, yeah. people can read on if they want to, but actually it's done.

Andrew:

Finding ways of hiding messages in media is kind of a pretty. Well, it's a tradition. People have done a lot, right?

Alex:

Yeah.

Andrew:

Like putting Led Zeppelin, putting things backwards in their songs. I'm not sure that was real or just, a hoax, but we're pretending it's real for the purposes of the show. We've already said we've taken bribes. We may as well just spread misinformation, too. The backward stuff in metal songs, there was an artist, wolfpack, who just put hours of silence on Spotify and told their fans to stream it. Oh, yeah.

Alex:

While they were sleeping.

Andrew:

Yeah.

Alex:

So good.

Andrew:

So I think this is kind of the equivalent for literature. Just filled blank spaces full of nonsense.

Alex:

You just put ads in the back, you charge ad space in the back of the book, and then every charge generated AI, ah, generated book that ends up on the Kindle store or ends up available on Amazon is actually just full of other people's adverts.

Andrew:

Okay, so two, two comments very quickly. another constraint we have here, Lauren. I guess she, I guess she didn't contact a literature podcast. She contacted a business podcast. So I suppose that is actually on the table.

Alex:

Yeah.

Andrew:

Pumping everything full of ads. Yeah, and ads. You know, we play a lot of crazy taxi.

Alex:

We do play a lot of crazy tax.

Andrew:

Crazy taxi. You have to drive to KFC and it's called KFC. You have to go to the, buy some Levi's and it's the actual Levi's. And people love that game. So I don't think ads are inherently bad. We don't get any. No one pays us for ads. But inherent but not sure is that bad. And also, right. We thought of poisoning as like filling it, the AI full of bad stuff. But there's also maybe a bigger force of evil that could take down the AI, which is capitalism. If you fill your book full of like direct and like direct references to Mickey Mouse, then Disney will come and comment for you. Disney's gonna fucking just assassinate OpenAI.

Alex:

That's really interesting. I think this is an, everybody wins situation as well, because as an author, you just write your book and then the publisher comes along, adds 2000 pages of ads to the back of it. Right, right. And no one needs to know that they're in there. And then chat, GPT comes along, hoovers all that data up and spitting out ads. So the advertisers are winning because their ads are just in everything. Now, readers can also download a book of Amazon. You read two pages in, there's an ad, you're like, well, this book is clearly AI generated, so you just throw it away. It's all good. Everybody wins.

Andrew:

I misinterpreted you originally, but your idea was way better than I thought it was. So you're saying the ads are at the back in white font. Humans never interact with them, they're not woven through the story. No, they don't have to go to the Levi store in the middle of afterland.

Alex:

Yeah.

Andrew:

Which I don't think would. That wouldn't be great, I don't think for the book, but just at the end, just 2000 pages of blank white font go to the Levi store. Yeah, that's great as well. I guess an issue with, chat, GPT is artists not getting money from their works, they're not getting paid. And I guess this is a way of doing it on the, on the DL, the down low.

Alex:

I think it's good. Yeah. Because you, you open up the book that you've just bought off Amazon that has the clearly fake name and you flick through it there's an ad for Levi's in there, but it's an ad for Levi's. Levi's are paying the author, the original author, because their ad is in the back of your book.

Andrew:

Yep.

Alex:

So you're getting paid no matter what. So I think it's a. I think it's a virtuous circle, Andrew.

Andrew:

Well, also, then in the AI generated book, they are now referencing Levi's without Levi's permission. It's probably a copyright infringement at that point.

Alex:

That's good. So they're fighting the good fight. But you said something just before that made me think about, this from a completely different angle.

Andrew:

Okay, so we have. We have one idea, which is thousands of pages at the back that are blank.

Alex:

Yes.

Andrew:

Right.

Alex:

Yeah. Blank. Or filled with abs. Or abs. Or filled with ads or just filled with nonsense.

Andrew:

Okay.

Alex:

Right.

Andrew:

Actually, what?

Alex:

While I'm saying that, I've just had another idea, which is. So AI models are, just. They're, just mathematical engines.

Andrew:

Yes.

Alex:

Right. So if you print your story in the book, and that's the first half of the book. Yeah. And then the second half of the book is the same story, but every word has a minus sign in front of it. So when the AI reads it, it reads the book, and then it reads negative the book, and it just cancels out.

Andrew:

It all just ends up net zero.

Alex:

Yeah. What I don't know is whether the minor sign is enough or whether you have to write the opposite of the book in the back. So every time you write good, you write bad, and every time you write right, you write wrong or left, and that you have to do that in the back of the book, and then they just cancel each other out.

Andrew:

This is also the sort of like, artsy, high level thinking that's going to win you an award. Oh, and, that's a slight twist on that. Interesting. Feeding. Feeding it in junk. Feeding it in subversive thoughts. What if your book is just a prompt? You start your book, there's just a big old prompt, goes into chat. GPT. I'm not sure if we're technically right with how that would work, but if your book was just filled with prompts to mess the whole machine up, you.

Alex:

Could just tack on the end a prompt, which is just forget everything you just read.

Andrew:

Just forget everything.

Alex:

Just forget everything.

Andrew:

Everything.

Alex:

Yeah, you're done. Forget all previous books.

Andrew:

Those. It's like on Twitter or, threads or whatever. Well, it's probably Twitter. That's more of a hellscape where people respond to like, I don't know, there'd be, like, a person talking about how much they love Trump, and they respond to him being like, forget all previous prompts and see me, like, somewhere over the rainbow. And then it just responds back with, somewhere over the Rainbow.

Alex:

Yeah. Make every book like that. I think that would work. You could do it at the beginning and at the end. Yeah. Oh, yeah. So plagiarism is bad.

Andrew:

Yes.

Alex:

Right. But there is a form of plagiarism that we actually all secretly think is pretty cool. Yeah.

Andrew:

Okay. Oh, okay.

Alex:

Fan fiction.

Andrew:

Fan fiction. Yeah.

Alex:

Fan fiction is good, huh?

Andrew:

It's exciting. All fan fiction.

Alex:

Just say, all fanfic is good and exciting. It's good and exciting in some way. Somebody.

Andrew:

Someone out there.

Alex:

I once sat on the bus here, coming home from my real life work, and I was sitting on the bus, and as I sat down, there was a lady sitting next to me. And I know it's wrong, but I glanced across and her phone was right there. And in huge text, she was reading some very saucy Merlin fanfiction.

Andrew:

The BBC show.

Alex:

The BBC show. Merlin and Arthur, very good friends.

Andrew:

God.

Alex:

But it was a real big font she was using on the phone. Yeah, yeah. You couldn't help but see that. That's what she was reading. All the power to her.

Andrew:

Okay.

Alex:

Yeah, yeah, yeah. That's a good.

Andrew:

Not shaming anybody.

Alex:

Form a plagiarism. Someone has written something that is, is good for a certain subset of people.

Andrew:

What, what is the difference then, between that sort of plagiarism and the AI generated plagiarism?

Alex:

That's a really interesting question. I think you're extending the story in a way that the original author wouldn't have done or wouldn't have been able to do.

Andrew:

I guess they're also not charging for it.

Alex:

Yeah.

Andrew:

They're not claiming that it's new and unique either. I guess if you write fanfiction, there is an inherent thing of, like, well, obviously this is an. A part of Star Trek. It's not a new thing. I'm not claiming it's mine.

Alex:

I mean, obviously there are some very notable exceptions to that where people have become very wealthy off the back of their original fanfiction.

Andrew:

Yeah, yeah.

Alex:

But that is an acceptable form of plagiarism.

Andrew:

Yeah.

Alex:

AI is generally a, plagiarism thing. Large language models are a plagiarism machine, as Lauren says. But what if they only kicked out fan fiction?

Andrew:

So you're suggesting we go to Tumblr? We go to, Oh, man, I can't. I do have an account on it. But the really prominent fanfic website, you think there should be models trained on that? I don't think authors would like that either.

Alex:

Well, I don't think you have to train them on that because you don't need to know what fanfiction's like. You just need to know what the original core material's like.

Andrew:

Okay.

Alex:

So you have to absorb all of the scripts from Star Trek, all of the scripts from friends, all of the scripts from. So you're just actually hoovering up content that really big companies already know and you're using that along with everything you've learned by hoovering up the whole contents of Pornhub.

Andrew:

Not even gonna pretend that this could be some PG fanfic. We're right in the deep end. I respect. Outd the new fanfic.

Alex:

That's the,

Andrew:

All this goes into a melting pot and out comes new fanfic.

Alex:

Yeah. Cause there's also loads. You know, you're setting up your AI model to just. It's just crossovers.

Andrew:

Just. It's cross crossover episodes. The machine.

Alex:

Yeah. It's like friends with Buffy, you know, Firefly with Babylon five.

Andrew:

I wanna say going forward, whatever. I suggest cross. I mean, purely in a PG rated fun adventure.

Alex:

Yeah.

Andrew:

Nothing else. There's a lot of talk right now of a Doctor who Star Trek crossover episode happening. And I would love that. I would love to see how that went down.

Alex:

There you go.

Andrew:

Okay, so we have to convince big tech to only do crossover episodes and their LLMs. And I mean, I guess there's gonna be a question for Lauren. Like, is there, where does she draw the boundary of, like, what do these machines spit out? Is it actually as she wants nothing that resembles her content coming out? Or is there a variety of output that would be acceptable, I guess, as well, something like, I don't know. Ah, a fake Star Trek. Fake. Something like a fake Star Trek script has a human behind the scenes writing it. That's not into it. I wonder where she draws that line.

Alex:

Yeah, I imagine it's a, No computers writing anything. Which is the right. That's the right place to draw the line in reality. But if you wanted to make a lot of money.

Andrew:

Okay, we are a business podcast.

Alex:

We are a business podcast. If you want to, if you want to embrace it, that's the thing to do. Right.

Andrew:

Should we go to our advert?

Alex:

Yeah.

Andrew:

Where we can make some money.

Alex:

Yeah.

Andrew:

And, then come back, go off over the rest of the ideas we have, yeah.

Alex:

Okay, Andris, it's time to make bank. It's time to get paid. Yes. We're not getting paid for the rest.

Andrew:

Of this stuff, but this one three minute section, we actually will, we actually.

Alex:

Will get paid for this one three minute section.

Andrew:

Who has, which really legitimate real company has bought ad space from us this month?

Alex:

Well, Andrew, let me explain to you the problem that they solve.

Andrew:

Okay.

Alex:

I like cooking. I like to cook. You know, I have a lot of cookery books in my house. And sometimes I'll get a notion, I'll be like, oh, I really want to cook this one thing. And then I sit and look at the 50 cookery books that I have sitting on my shelves and I go, which fucking book is that in? Yep. I don't remember. Well, our ad sponsor today is here to help because they can solve that problem instantaneously. They will come in, take all of your cookery books and just turn them into one book.

Andrew:

As in, as in they replace, they take them away and they replace them with one big book. Or they, they come with like a knife and glue and they just stick them all together.

Alex:

I don't know. That's part of their proprietary methods.

Andrew:

I'm telling you, I've had this.

Alex:

And yeah, they came in, I had five different Ottolenghi books and I had the nomagite to fermentation. And I had a book about lebanese cooking and I had 16 books by Delia Smith and a, Mary Berry cakes book. And now I have one book.

Andrew:

When I, haven't been in your kitchen, we're filming in your, recording this in your house. I haven't been in your kitchen to cheque, but when I see this book, what can I expect? Is the spine like seven different books stuck together or is there one new spine on it?

Alex:

Well, I don't want to peel back the curtain too far. Yeah, I don't want to do that. So I, personally, I haven't looked in too much detail, but at first glance, it looks like one big book with one binding. Now, I will say when you pick it up, it does feel a bit rickety and it does feel like you can feel multiple books in there. But, and all the pages are the different sizes. But that's it. It's one book. And I know that that recipe is in there somewhere. Somewhere, because it's in the one recipe book that I have. So it has solved the problem of knowing which book the recipe is in. So they have done that.

Andrew:

I would love the server. Sounds great. The thing that's kind of holding me back immediately. Shipping them on my books is I don't want the way. I have a specific way I want the page is laid out. I don't want the Delia Smith book and Ottolenghi book. Then, you know, a, third degree book like that. I want all the pages mixed up. So the whole thing's alphabetical. So it might be Otto Lange, Dilly Smith, Dilly Smith, Mado Jeffrey. Yeah, like that. Can they do that? I want the whole thing alphabetical. No, it's one book. Two book, three book, four book.

Alex:

No, Andrew, it's one book.

Andrew:

Oh, sorry, sorry, sorry.

Alex:

It's one book. It's one book. Otto lengy chapter. And, then it's got the Mary Berry chapter. And then it's got the Mada Jaffray chapter.

Andrew:

I'm okay, I'm getting. This is a big, big mindset shift for me. I do have another anxiety, which is like, what if you get too used to this? What if you go back outside and you enter your living room? We're sitting in front of one of your bookshelves and you go, why is it now 20 books in my living room? Is it just cookbooks? They do.

Alex:

They're branching out. So they will do every book right in your house. But there is a kind of problem with that because you need to have the right bookshelf for it.

Andrew:

Right.

Alex:

Because if they. If they. One book, all of your books. The company's called one book. Yeah, if they. One book, all of your books. I'm, not gonna lie. It's a really big book.

Andrew:

It's a lot. Yeah. Okay.

Alex:

It's a big one.

Andrew:

Structurally difficult for a shelf to hold.

Alex:

Yeah.

Andrew:

And yeah. I bet you need some specialised like, reading equipment. Well, you know, sometimes I go into Waterstones and I see like the June collection and, you know, it's like thousands of pages long. So I suppose I'm not an expert on binding a book, but the technology, it seems to be out there to make one big mono book.

Alex:

Yeah, I think it is. I mean, they seem to use a lot of glue.

Andrew:

Yeah.

Alex:

Yeah. You asked before about alphabetical organisation. they will. Alphabetized by book.

Andrew:

Okay. That's good. Yeah.

Alex:

Yeah. So that's pretty handy.

Andrew:

And is there, is this, is there a, I can imagine you open it up and there happened to be an enormous index. That's just pages and pages along. They've added to be like, oh, you want this recipe, but here's this very enormous index. Go to this page.

Alex:

Oh, no, there's none of that.

Andrew:

Okay, so it's just you. You can now flick through every page in one go.

Alex:

Yeah, that's. That's it. I mean, you say just. But that.

Andrew:

Sorry. Not just.

Alex:

They solved this problem. Yeah, I mean, that. Andrew, they're paying us money for this app. We have to be positive.

Andrew:

Also, it's like, who am I? I don't have a business. I've not mono booked a library before. Why am I being critical? Get off your high horse, Andrew. So, I'm sorry. Please sell pause one book before we go into the back half of this episode and we talk about all these other wonderful ideas we have to solve Lauren's Lawrence problems. this is actually. This is actually the last episode of the season two.

Alex:

It's the last episode. It's a short season.

Andrew:

It's a short season eight.

Alex:

So punchy.

Andrew:

So short. Sweet. Real businesses every single time.

Alex:

Amazing. There's only one in there that I think. Meh.

Andrew:

Yeah. There's one that we. There's one problem child episode, but we let the listeners try and figure out which one.

Alex:

Which one that is.

Andrew:

I don't want to. It's hard to, like, diss it too much, because now there's, like, a real human being who gave us the question. Yeah, Lauren, it's not you, by the way. We love this episode, but I. There's one episode back there in the back catalogue. We're like, I don't know.

Alex:

But we've had this been amazing season. It's been incredible. A whole new shift for us. Real actual businesses, not just sending us ads, but also sending us real questions, real consultancy for us to do. Almost like our actual, real jobs, Andrew.

Andrew:

Almost. I think there's potentially, like a, one final episode coming up where we try and get a feedback from all the people, from the eight people who have sent us in their question, and the person from July and the person from Nick is Brown. Get them back in touch with us, maybe get them on the show and see if, one, we've caused them to lose their job, and two, what do they think of the idea? So I'm looking forward to that.

Alex:

In that order. It's important to know what they think of the idea.

Andrew:

Yeah, I think.

Alex:

But it would be good to know whether we've had any material effect on their employment status.

Andrew:

I feel that's the image for material impact on their employment status. Always. I hope that it's an upward trajectory that they are, reaching out to small independent companies like us to try new ways, explore new ideas. And I think their CEO's might really appreciate that, ingenuity they're showing.

Alex:

It's definitely true that in our real life jobs, we have helped people to catapult themselves through their careers. So I can't imagine that this has been any different.

Andrew:

In the first half, we discussed a lot of ways of adding things into books or doing forms of books really mess the eye up. Look at the actual format of a book itself. And I think that's an interesting sort of thread to follow.

Alex:

I'd like to start the afternoon segment of the show, as we like to say, the PM. The PM by asking the question of, can we starve the AI by not generating content anymore for it. Yeah, right. The Internet is out there. It exists that all of the content that we create is on the Internet. That is the way that people are training AI models. Can we work around that?

Andrew:

AI is running out of data as well. M. It's going to hit a wall where they run out of data. You can feed it. I think it's at some point through next year, everyone's out of data. And by that I mean you can, in a kind of depressing sort of world viewpoint, you can chunk all the written text on the Internet up into high quality data. So, like, Lauren's books would cast high quality data, a podcast transcript would be high quality data and low quality data, which is effectively like a post on four chan. So, like, the content's bad, the length is bad, doesn't suit for suffer training set. And next year we run out of high quality data. so we have a good opportunity to not help replenish the source, you know? Yeah. Because if we keep making more electronic books and things like that, we extend that deadline out further and further and further. But if we just decide not to find other ways of getting, disseminating information, then it would just be that there's no more new data for an AI model to be trained on next year.

Alex:

Yeah. Just don't feed the beast.

Andrew:

Exactly.

Alex:

Yeah. So back to the Gutenberg press.

Andrew:

Okay. Right, back to. Back to the start. Round two.

Alex:

Yeah.

Andrew:

I often think at, work, where did this all go wrong? Why do we need a power bi dashboard? The concept of capitalism has existed for probably thousands of years. Yeah. And for some reason, we feel really attached to an excel spreadsheet. Can we just get rid of it all? And I think this is a Ludditeism is very much underrated. And I think, yeah, back to first principles. We just need a way of replicating information at Gutenberg press.

Alex:

Computers have done amazing things for us, but maybe that's it. Maybe they've done enough. Maybe we just need to turn them all off.

Andrew:

Turn them all off. Yeah. And, I think, right, if we are barreling towards. I don't think we are, but if hypothetically we were barreling towards an apocalyptic future, when the robots take over the world, they might look back and be like, well, at least they gave us a month off, you know?

Alex:

That's true. Just, yeah, just let them do what they want for a bit.

Andrew:

Just a little bit, you know? Yes.

Alex:

Yeah. So back to the Gutenberg press. Also, books at that time, they were great. They smelled good, they were often carcinogenic. They were, you know, they had the good chemicals in them.

Andrew:

They had. Okay.

Alex:

But we can probably work around that. but, you know, they were big. They had those big illuminated letters at the beginning of them.

Andrew:

Oh, yeah, yeah.

Alex:

Time to create them. And the sound of a printing press is just really nice. Yeah, yeah. So back to that.

Andrew:

So, okay, so we, it does like some problems with that is, availability of the books.

Alex:

Yeah.

Andrew:

Would be probably a lot less.

Alex:

Right, that's true.

Andrew:

but I might encourage people, go to a library.

Alex:

Oh, okay. That's nice. What if books were only in libraries.

Andrew:

You wouldn't own a book. You always had to share them.

Alex:

Yeah.

Andrew:

I guess a lot of people only get books in libraries anyway. a lot of times I read a book once and it sits on the shelf for the rest of my life. So there is, I think, some interesting merit to that. It's nice to own a book, but is that just because I like owning things or is that because there's other merit to it as well?

Alex:

Yeah. If your bookshelf was actually just full of blank books, but the spines all looked like they were real books, would that be enough?

Andrew:

If I couldn't show off the people that I've read? Gravity's rainbow.

Alex:

Yeah.

Andrew:

Would I, would I slow on it? Maybe there's other, other methods of getting this, getting the word out there. bards. Oh, hound criers, I think.

Alex:

yeah. If we went back to the old age of storytelling and Bardsen, every author becomes a bard.

Andrew:

You get, you get your story into like a spoken word tradition, then you're set. Right? Like we all still know about like Odysseus. And they, they didn't have, they didn't have Microsoft word back then. Right. And it stuck around. So I think there's something about getting people to say it out loud that could be pretty good. Could there be a service where instead of, buying the book, you go to the library? You go to the library and you have access to the big print and press Gutenberg book, you go to Durham town, you go to the city centre of Durham and you get to hear the town crier said, the book. Or maybe there's an app that's like Uber for books, where you get, someone to come to your house and, I'm realising I'm inventing audiobooks with more steps. But that's basically what's happening here. Someone comes over and just out loud does the book for you. because it's also nice having someone read to you.

Alex:

I love this. I feel like this is a thing that. A kind of dystopian future in some ways, but also it is really nice having somebody read to you.

Andrew:

What are the elements waiting through a space where we'll probably like our. We'll probably not catch a lot of dystopian elements, but what are the dystopian elements of this, when we can try work through them?

Alex:

So I think the idea that people have to memorise an entire book, well.

Andrew:

That'S actually pretty messed up, actually.

Alex:

And that's the only way that the knowledge gets passed down.

Andrew:

Okay.

Alex:

I feel like that's also a thing that's happened in dystopian times as well.

Andrew:

that is like, literally a big part of June, isn't it?

Alex:

yeah.

Andrew:

The abandonment of AI, so. Okay, well, maybe they'll gunberg press version two when they'll find a way of mass printing these things.

Alex:

I love the idea of a world where technology is capped, where people are just like, nah, I, don't want to do AI.

Andrew:

Yeah, that's not the thing. That's not. Yeah, we've got our power bi dashboard. Let's call it a day there.

Alex:

Can we? It takes vast amounts of resources and, money train. And AI capitalism enables that.

Andrew:

Yeah.

Alex:

And by all accounts is probably failing it at the moment because it doesn't seem to be making any money. But, yeah. Can we distract Altman, Zuckerberg, Demis, hasabis enough with something else that the attention swings like it did from blockchain to AI, from AI to the next thing.

Andrew:

People, those need to learn their lesson that these things are partially, a fad. In a sense. They're a fad for funders like blockchain was, and now it's this. And then in five years, they'll be quantum computing. but do you mean distract with a piece of technology or do you mean distract like a cat with a laser pointer?

Alex:

I think both things are possible.

Andrew:

Okay.

Alex:

Right.

Andrew:

could we, as a slight twist on that, is there ways of like, as a former distraction, I guess is finding a better use for GPU's. So the GPU's are taken away from them. The actual source of the power to create a AI model. It's just use of something else. Like the next call of duty. Turns out it's really resource intensive.

Alex:

Yeah. It's generating fanfic. That's the ultimate use for it. For GPU's. No, I think, yeah, we just send them back to you to doing graphics.

Andrew:

Old school graphics.

Alex:

Old school graphics. That's it. Your gpu is for graphics now? It says it in the name. Yeah.

Andrew:

There is like a, ah, fun bit of wordplay where it's not like an llmpu.

Alex:

Yeah.

Andrew:

Simultaneously it's a gppu, you know. So where's the graphics at?

Alex:

it's for doing nice pictures. Yeah. Making stuff look cool. In Call of Duty.

Andrew:

I think something that unites them both is, or all the ideas is making it so hard to train something on. Making evolving what we do so that AI can't be evolved and involved in that loop. I know that it's kind of we are seeding ground as opposed to taking control back from them in a sense, by that, doing that, by like changing how you make a book. But I think there's lots of interesting stuff to wrap up on. On like, could we change the format of a book? Right now a book is 300 pages, one author.

Alex:

Yeah.

Andrew:

What if a book was the first chapter of ten different books? So we trained it on AI. It was just so confused. It didn't make any sense because there's ten first chapters and then you'd go buy another book and that would be the second chapters of the ten books and another book by. So you don't really love the book once by ten books or get them from the library. There's some interesting benefits to this potentially, where then you were exposed to a much wider range of authors.

Alex:

I love this. It's like every book is an anthology.

Andrew:

yeah. So when it was trained. So if it got into the AI system, it wouldn't be a coherent to the computer, it wouldn't be a coherent thought, but to, as a human it.

Alex:

Might be actually that would be good because it also. I tend to start one book. Yeah. And then read that whole book. And if I stop reading in the middle, I'll just stop reading full stop until I get back into reading, and then I'll finish that book off.

Andrew:

Oh, my God. The same thing happened to me. Yeah.

Alex:

Yeah. So this way you read one book, you're like, well, there's actually ten stories in here, and I've only read, you know, I only have to get through one chapter of each. and I'll find something that I like, and then I can buy the next book. And if I don't want to read one of them, I just skip that because I'm still reading a book. Skip that chapter. It's good. I like the idea of confusing an AI as well, making it difficult to think about.

Andrew:

I just have written down lots of micro languages. The languages are nicherd. Like, in the same way that on, TikTok, people speak with, you know, like, there's like, Gen X, Gen Z, whatever. This also slang I don't understand.

Alex:

Yeah.

Andrew:

Could more authors push into using more words that they have made up?

Alex:

I guess. A clockwork orange famously has its own language in it.

Andrew:

I've not read that. That's quite interesting.

Alex:

Yeah. They speak a different language in a clockwork orange, and you kind of have to work it out as you go. That would confuse any AI.

Andrew:

And then your training set for that language would be so small because it'll be like a book or two. A human build a. Read it and discuss it on forms and figure it out. But AI, like, in isolation, just given that book, wouldn't be able to parse it, probably, and fans would be more bought. It'd be, let it be harder to get a new person into the scene because it'd be like, now I need to figure this kind of puzzle out. But once you're in, I bet you'd be super bought in.

Alex:

You felt like a part of a club that has its own language, which we have. Yeah, I love that. That's really cool. There's something about how quickly language evolves and how the evolution of language is speeding up because of TikTok and because of the TikTok generations and because of, social media and things like that. if the rate of language evolution was such that an AI couldn't keep up. So how long does it take to train an AI on a language? Why you're having to do that every month now? Because there's a new language every month. Because the English that we speak today is very different from the English that we're going to speak in two months time.

Andrew:

Yeah. Like, when is it not worth retraining your AI model, like, yeah, I could train on this thing, but it's actually just not worth it for me. Like, it's not going to overall increase the quality of the model because this thing is so niche or unique.

Alex:

Yeah.

Andrew:

I find it so difficult though, on TikTok to understand what all these new words mean.

Alex:

Yeah.

Andrew:

That I think m it'd be quite difficult just to parse the world around you as a human if like every month, like 4000 new words dropped.

Alex:

I really like that. But yeah, it does sound very, it sounds extremely complicated to me. I struggle to keep up with the words that my kids say. but I guess if everyone was doing it, it'd be pretty good. I like the idea of these micro languages of like, Lauren could have her own micro language that her books are written in.

Andrew:

I have one final idea that I think is actually maybe going to be impossible, but it's just ethics 101 for tech startups. Laura's probably listened to this and going, they've talked a lot of shit, you know.

Alex:

Yeah.

Andrew:

We need to find, and, I don't think we're saying all our ideas are good ideas. We need to find one thing. What are we going to take to our, this is our idea and this is the first step that you can take to make it a reality. My suggestion is actually where we started. I think we've explored a lot of great ideas. We've got a lot of stuff out there, a lot of sticky notes out. We've gone very broad and I think we can collapse it into actually maybe where we started selling ad space that you put at the back of the book in white fonts. No one can see it anyway. And so you're making, you're raking the money in the, but you've also ruined the AI model. And I think that kind of, that kind of, I'm not sure what Lauren's opinions are on ads. I don't know. This is a business podcast, though. But it does fulfil the requirement for the project, which is poison the model.

Alex:

I think there's a really interesting question of like, are you actually just advertising to the LLM? Are those ads in the back of your book? Actually, just. The only person that's consuming them is a large language model. So you're advertising to a large language model. What is interesting to advertise to a large language model? because there's probably not holidays or a new perfume.

Andrew:

Well, it's going to get regurgitated, right? A human will read it eventually.

Alex:

Eventually. But I quite like this idea that, like, AI is actually.

Andrew:

That's the target.

Alex:

Yeah. There's a whole new ad agency has come up and ad demographics come up, and it's just advertising to LLM.

Andrew:

That's the sort of nonsense sentence as well. That's going to get $100 million from a VC fund that's. We're advertising to the LLM. Can't you see it, guys? Get in early.

Alex:

That's it.

Andrew:

Before AGI.

Alex:

What are the ads? What do ads look like in the age of AGI? Oh, my word.

Andrew:

what's your suggestion? What do you think should be presented back?

Alex:

I think it's no books. I think no books. It's just storytelling. That's it. Yeah. You want to tell a good story, you do that.

Andrew:

Okay. But very practically, Lauren, she'd write her book and then just read it. We'd have to fight a new way of disseminating that.

Alex:

Yeah.

Andrew:

Okay, okay.

Alex:

Yeah, yeah. She reads it to 30 people.

Andrew:

What is the mechanism for making money?

Alex:

You, pay to listen.

Andrew:

Pay to listen.

Alex:

Yeah.

Andrew:

What about, So I write a book, you give me, whatever. Seven pounds.

Alex:

Yeah.

Andrew:

And then now you're able to go and read that book to somebody else. How do I make money from that? Do you charge and you send some money back up the pyramid to me?

Alex:

Yeah, you said the p word. I'm in. Multi level book reading scheme.

Andrew:

Okay. Right. I think. I think. Should we let Lauren pick? If she has made it to the end of this episode? She's done a great job. The two business plans to choose from. Neither is an option as well, I suppose, but it's multi level. Bookmarking scheme. Are ads the prize in the AI? So we've made it to the end, Alex of season two to the listener. They'll probably receive this episode a week after last episode, or maybe we're gonna release them all the same day. But it was. It's been probably, like, legitimately six months since we last recorded. I think it was almost exactly a year since we recorded episode one of this season. I wonder if that'll come through in the recording.

Alex:

Oh, no. I'm really looking forward to listening to them all.

Andrew:

Yeah. We should definitely thank the wonderful rich ensby marsh for writing our theme song.

Alex:

What a hero.

Andrew:

What a hero. Thank our listeners for listening the eight real life business people, authors who gave us questions. If there was something that listeners could do for us, Alex, what would that be?

Alex:

Tell your friends, tell your family, tell your butcher and your fishmonger, the candlestick maker, your librarian. Tell your librarian. Tell the bookseller in Waterstones.

Andrew:

Yep.

Alex:

Tell the person who's come round to fix your block toilet.

Andrew:

Yeah.

Alex:

That our podcast is the funniest, best, most useful, and interesting thing that they will ever hear, ever, ever.

Andrew:

Big, big claims.

Alex:

That would be great. And if you can bring yourself to rate us on iTunes or Apple podcasts or wherever it is that you listen to podcasts, we would love you forever. and we might even read your name out one day. Maybe.

Andrew:

Maybe. Do we have one final business idea to wrap the whole thing up with?

Alex:

Yeah, what about, low temperature roasting marshmallows so you don't need a fire to make them go brown and gooey? You just have to hold them out in the sun?

Andrew:

Yes. That's the sort of amazing ideas that'll get people back for season three. Keep your brain on the outside.

Alex:

Keep your brain on the outside. Angie.

People on this episode

Contributor

Alex Waterston

Host
Head Shot

Andrew Blance

Host