Co-writing Fiction With Generative AI With Charlene Putney

How can authors use generative AI as a co-writing tool? How can creatives approach AI possibilities with curiosity rather than fear? Charlene Putney talks about writing with LAIKA.

In the intro, ChatGPT, thoughts on the GitHub Co-Pilot case [WIRED]; and why digital abundance is an opportunity for curious creatives, not a threat. I also mention my book, Artificial Intelligence, Blockchain, and Virtual Worlds: The Impact of Converging Technologies on Authors and the Publishing Industry, and you can find more interviews and resources on my Future of Creativity page.

This podcast is sponsored by Written Word Media, which makes book marketing a breeze by offering quick, easy and effective ways for authors to promote their books. You can also subscribe to the Written Word Media email newsletter for book marketing tips.

Charlene Putney is an award-winning writer, professional speaker and university lecturer. She's also the co-creator of LAIKA, an AI-powered creativity tool.

You can listen above or on your favorite podcast app or read the notes and links below. Here are the highlights and the full transcript is below.

Show Notes

• Why people with arts backgrounds need to get more involved in AI
• AI-powered creativity tools for writers
• Using your own backlist to train an AI brain
• Tips for approaching AI tools as an author
• Copyright, bias, and plagiarism
• Data licensing and how fine-tuning models might benefit creatives
• The importance of personal branding, author voice, and connection with readers in a future of digital abundance

You can find Charlene Putney at AlphaChar.com and on Twitter @alphachar. You can try LAIKA at WriteWithLAIKA.com

Header image by Joanna Penn on Midjourney.

Transcript of Interview with Charlene Putney

Joanna: Charlene Putney is an award-winning writer, professional speaker and university lecturer. She's also the co-creator of LAIKA, an AI-powered creativity tool. So welcome to the show, Charlene.

Charlene: Thanks so much, Joanna. It's really great to be here. Thanks for inviting me.

Joanna: Oh, I'm excited to talk to you about this. So first up –

Tell us a bit more about your background at the intersection of creativity and technology.

Charlene: Sure, so I am quite old now. I'm 42, so I've been around doing a lot of things for a long time. I started out my “career” – I'm doing inverted commas here – studying ancient Near Eastern languages in university.

And then after that, I wasn't quite sure what to do next, but then they had openings for people to work in Google in Dublin, where I'm from. And I applied there, and I worked there for the next almost five years. And then from there, I went and worked in Facebook.

And weirdly, my ancient Near Eastern languages helped me out a lot because I was able to use those to work on product teams with right to left languages. So I worked in tech there for about almost 10 years.

Then I really wanted to do something a bit more creative with my time. And I had been doing a lot of experimental writing classes and I was writing a little sci-fi novel myself. And so I basically left Facebook to just devote myself for a few months to writing.

After those few months had gone and I realized that I still needed to earn money and pay the rent, I started trying to figure out how I could actually use my writing to do that. And that's how I got into writing for video games.

So after a few small projects in and around Dublin, I ended up working for Larian Studios on much bigger games, big role-playing games, like Divinity: Original Sin 2 and Baldur's Gate 3.

And from there, I started speaking at different games events all around the world, and one of them was here in Copenhagen, where I live now. And there I met Martin Pichlmair, who's my partner on LAIKA and my partner in life. And together, we've basically been just making wild experiments with writing ever since. So that's where I am.

Joanna: That's awesome. And just to say, I'm older than you. So I'm 47.

Charlene: Oh nice. I never meet people older than me.

Joanna: But also my degree is in theology, and so I did ancient Greek, and studied Israel before the exile, and some ancient Near Eastern stuff and I started out doing Arabic. So I kind of feel like we have quite a lot in common in our background, but not computer science, right? And I think this is so important. Let me then ask you that question, essentially you've got an arts background –

How does your arts background help you in this technological world full of people who do computer science?

Charlene: That's a really good question. So in my undergrad, I also did philosophy. And I think that part of what has helped me a lot, like in all of the tech world and also in the games industry, and now also coming into the AI industry where I've been for the last year or so, is this feeling of, it's okay to stop and slow down and think about things before rushing into solutions.

I've kind of always been the person in the room who's going, okay, wait a minute, let's just think about this other part first before we implement something.

And I'm sure you can see now, especially with the ongoing downfall of Twitter that we're all witnessing every day, that is something that is not super common in the tech world. So I feel like that's one of the things that arts has helped me to bring into my career.

Joanna: Yes, well, we're recording this in the middle of November 2022. So by the time people listen to this, we don't know what might have happened with Twitter. But it's interesting, just staying on the creativity side, so I feel pretty enmeshed in the AI space as well.

And I often feel like there aren't enough – when I say creative, I mean, obviously coding is very creative. So there are lots of coding creative people in the room, but there aren't so many, let's say ‘artists'.

So you mentioned writing a science fiction novel, and I feel like the visual art at the moment is obviously really interesting with AI. But this sense of ‘are we missing artists and arts people in this tech space'?

And how can arts people get involved when they feel like the AI space is just too techie?

Charlene: Yeah, I think that's a really nice point, Joanna. So actually, like when we started making LAIKA, it was kind of coming from that place of, my partner is a programmer, like he has a computer science degree. And so when we would make our experiments, he would set up all these notebooks for me in Jupiter and Collaboratory, and he would write code programs for me to mess around with and then I'd be able to work with them because I don't have a coding background.

What we really wanted to do was make this accessible to normal people who don't want to fiddle around with knobs and buttons and find their way through that.

And that's one of the things I've really found very strange about so much of the tools out there, is that like even for example, Midjourney, which is my favorite of the image generation tools and I use it all the time for illustrating my yoga workshops, illustrating pieces from LAIKA, is that like it’s in this Discord server, and you also need to understand how to make the prompts. And all of the things about it just requires so much personal investigation and knowledge about how to do things.

Like even just being on the Discord server, I tried to keep my dad interested because he's really into lots of different types of tech, and then the Discord thing, he just couldn't get into the Discord part.

So there is that, I think there is just a barrier of entry to regular people like me, still at the moment, even with these excellent tools. So yeah, I hope that that becomes more accessible. And I'm sure it will because every tool, when it comes out at first, the early adopters need to jump through a few hoops and then people make it more accessible for the next ones who come along.

Joanna: I totally agree with you. And I mean, what's so funny is I had gone on to Discord for various blockchain things, and I just hated it. And I was like, I'm not even going to use Discord, I'm just not going to go there. And then with Midjourney, I was like, well, I need to have a look at Midjourney. And now I'm in it every single day.

Charlene: Yeah, same.

Joanna: I will get over this problem with myself and Discord in order to use the tool. But I mean, it is a really interesting time. But let's talk about LAIKA –  

Can you explain what is LAIKA, and why did you decide to create it?

Charlene: Sure. So it's basically an AI-powered creativity tool for writers. It's specifically for like serious writers, we would say. So like professional writers, really serious hobbyist writers, people who have a chunk of writing already behind them basically, like who already have their own voice and their own way of writing.

The way it came about was actually over the pandemic, myself and my partner would be sitting around the table together because we both be at home from work, working from home. And so we had a lot more time together to talk about our experiments and think about what we want to do next.

And we made tons of experiments using GPT-2, using Google Colaboratory. And then we started thinking, “wouldn't it be so nice to show other writers how to do this?”

So we started setting up these events online, like over Zoom, where we would show people, over a two and a half to three hour session, very, very step by step on how to upload their texts, how to train an artificial intelligence on their own writing, how to use these notebooks, and then how to decipher the results and pull them into something else.

And so we got a great result from those and you could just see the magic in their eyes, the sparkle. But it was still spending three hours on a Wednesday evening trying to learn this stuff and then trying to hold the way that the code notebook works in your brain before you try and work on it again.

So basically, we were just like, well, maybe we could make this very, very accessible. Like maybe we can take this and put it into a very friendly, intuitive and accessible format. And so that's what we basically did.

So when you come to work with LAIKA now, you can train a brain, which is basically like fine-tuning an artificial intelligence using your own writing.

So as I said, it's for writers who have a body of work.

So you can basically come in, give us somewhere over 10,000 words of your own writing, we will fine tune your own personal GPT-2 model on that – we host our own GPT-2 on our own server – and then that'll always be there for you.

And it will basically bounce back suggestions and ideas and thoughts to you, in your own voice using your own concepts. So that's what the idea is. That's where we've been going with so far.

Joanna: It's so interesting. Obviously, I've tried loads of these different tools. And in fact, I did work with someone to train a GPT-2 model on my writing before I'd discovered yours. So I think this is a really interesting thing. But what I actually found by doing that is that I was kind of bored by my own brain.

So what I did, like you mentioned this is an AI-powered creativity tool, and what I have kind of come to is that I initially thought I wanted more of my writing, but actually, I love the creative copilot or almost the sort of madness that having a GPT-3 or whatever brings because it's the fact that it's not my mind that is more interesting than being my mind.

So talk about the different brains in LAIKA you can use other than your own work.

Charlene: Sure. So yeah, that's also a really nice point. Because indeed, sometimes our own voice is not as inspiring as that of others.

At the moment, we've uploaded a bunch of different brains into LAIKA, which are all based on works that I've taken from Project Gutenberg. So they're all in the public domain.

So we have like, for example, Bram Stoker's Dracula, Mary Shelley's Frankenstein, we have Kafka, Dostoyevsky, Jane Austen, and you can basically use all of these. But our best one is Marcus Aurelius by far, he is just so good. No matter what you do with Marcus, he gives you a great answer. So I'm kind of obsessed with that. I will use my own brain, but then I will always find out what Marcus has to say about the same thing and sometimes take that instead because he's my favorite writing partner now.

So yes, we found that we do a lot of live events at the moment. So like, we were just at the Royal Danish Academy the other night, and we were at the AdventureX conference in London last weekend. And when we're at these events, we basically do a live demo, where we use lots of different brains to kind of create a story or create a thing. Like for AdventureX, we made The Five Rules of Writing with LAIKA, which you can see on our ‘Write with LAIKA' Twitter account.

But basically, there's so much joy in the audience, so much joy in the moment of like bouncing between different brains, and like starting one sentence with the mind of Robespierre and then finishing it up with H.P. Lovecraft. It's wild.

And actually, another thing we've been doing is making combination brains.

So myself and my partner are actually writing a little book at the moment together with a brain that we've made as a combination of Jane Austen, H.P. Lovecraft and a lot of really crappy free cryptocurrency eBooks. And we're making this like cosmic horror marriage plot about demons on the blockchain. And I think it's gonna be great.

Joanna: That sounds awesome. I'm definitely interested in that. I often have demons in stuff I write.

So there's a few things to come back on there.

So you talked there about joy — we're both giggling away here — you've talked about magic, and sparkle and wild.

And these are all words that make this sound amazing. And having used many of these models, I spend a lot of time being amazed and finding it super brilliant, but most writers are just scared and rejecting this out of hand.

I've had, I'm sure you have too, but quite a lot of personal attacks and comments and emails that say, “Why are you talking about this? This is going to destroy writers,” and all of this.

So what do you say to people who are scared about it? How should people approach these tools?

Charlene: So yeah, another great question, Joanna. And the funny thing is, I don't get much vitriol, actually, at the moment. Fingers crossed, it doesn't start.

Joanna: You might now, coming on the show!

Charlene: But I suppose, I find that if a writer spends any amount of time writing with artificial intelligence, it's very, very clear, very, very quickly how much the human soul and creativity and spark is needed to kind of corral thoughts together and pull the pieces into a beautiful tapestry that the world will see as good writing.

That without you, the writer, pulling something into it and taking it together, it isn't anything. And it never is. And I don't think it ever will be.

I mean, maybe I'll be wrong, maybe there'll be amazing AI poets in 50 or 100 years.

I feel like as soon as you start working with it, you start to see it's a paintbrush.

It's a tool.

It's like when we moved from writing with a pen to writing with a typewriter. It's like when you move from writing with a typewriter to writing in Google Docs. And now here, we have this thing to bounce off.

So I suppose my experience is that like, I used to work in a very big writers room, there was seven of us. We all worked in the same Google Docs, we wrote over each other, we added comments to each other, we struck out each other's lines and put in new ones. And it was just this constant like hive mind way of working. 

Then when I left the writers room, when I moved here to Copenhagen and was back writing by myself again, I was completely confronted with that fear of the blank page. Honestly, I hadn't really ever experienced that before to the same degree, because I completely missed having somebody to bounce off. And for me, that's what AI writing tools are, all of them.

It's like, they're not replacing me, they're not replacing anything, what they are just doing is giving me something to bounce off. So I can be like, “Oh, yes, more like this,” or “Oh, no, I don't like that at all.” But even that's giving me a new direction to go in.

So it's kinda like all those tricks that the old experimental writers used to work with, like cut-up technique, like the Dice Man method, all those things that the Surrealists used to try to pull out stuff from the unconscious. That's what we're getting.

Joanna: It's almost like you prompt the AI with something, and it returns something back, and then that prompts your mind to come up with something else, and then you come up with a new thing and then re-prompt it. It's this iterative process. But to me, the writer, the artist, is the creative director. 

Let's get into how writers make a living right now. We make money from licensing copyright, that's how we make money. So there are two things to address here.

First of all, the issue of creating these models in the first place. So obviously these tools have been trained. I mean, you said you trained LAIKA on out of copyright works, but GPT-2 itself, and GPT-3, and Midjourney and all these have been trained on works available on the internet. And I just don't think we can say they are all out of copyright. I just can't see it for text, as well as images.

How do we address the issue of the training data for models, given that the original creators are not being recompensed?

Charlene: Hmm, okay. Yeah, that's a really juicy and spicy question. So I suppose another part of that is that the bigger the models get, the more of the content they're going to be taking from the world.

So as far as we know, when we were working with GPT-2, we only work with GPT-2 at the moment because we want to be able to host our own models ourselves, so that we can give our users their own brains and have it in their own space.

But with GPT-2 is that it was like trained on lots and lots and lots and lots of scraped internet data from around 2017, and a load of old books, I think called ‘The Pile'. And a lot, a lot of the content comes from things like Reddit posts, social media, anything that's out there and that's publicly available.

So when it's being recombined in that way, that content, like the content of GPT-2 itself, doesn't really give you anything special. So if you try to write something with it, like you can't maybe write something really amazing with it. There's nothing that you're going to create out of it that somebody will be able to say, “oh, that's using my content,” or “that's doing something that was mine.” Whereas when you fine tune the work, then it is like honing.

So I like to think of it like GPT-2 is basically the grammar. It knows where the words go, it knows how a sentence is constructed, it knows what the shape of the sentence or a phrase should be. And then we're filtering that through a particular writer's voice.

And like we're very, very interested in following the law, and not in any way interested in being any kind of proto case for the new laws that are definitely going to be coming in the next few years around AI and copyright.

So we are very strongly saying people should only use either your own content or content that you know to be in the public domain.

And if anyone uploads content that is in copyright, I will take it down. So we have that and like, yeah, from our perspective, the GPT-2 is open source and we're just using it as a base grammar.

However, you are kind of inching me into one of my – what's the opposite of a pet peeve?

Joanna: Favorite topic?

Charlene: Favorite topic, yeah, which is this compensation for the writers. Because one of the things we're really hoping to do in our next few months, like as we get ourselves with a brand, we've started doing this thing called shared brains.

So you could basically, Joanna, take your brain, and upload it and share it with other people for them to use.

What we want to do, is allow this to be a kind of a thing where IP holders can license out their brains.

So let's, for example, say my dad's favorite author is Robert Ludlum, who has been dead for many years and yet still keeps writing books with the aid of ghost writers and additional writers.

And how wonderful would it be if the estate of Robert Ludlum was to give us all of his content and then licensed that to us so that people who want to write just like Robert Ludlum would be able to pay a small fee that would go almost completely to the IP holder in order to access that, or that you or any other writer who has fans, who has people who want to write like them, can license their brains for a fee.

Joanna: I love that you say that because two years ago, I wrote a book on AI and how it will impact publishing and authors and talked about this.

I basically said, what I want is to collectively license blocks of IP.

So if I said, for example, indie thriller authors. So indie authors control all of their rights. So this would be very hard with the publishing companies, probably, because they're quite resistant.

But there's quite a lot of indie authors who would join together and create like, let's say, an action/adventure/thriller brain that would have the work of quite a lot of my friends.

And why I put it with Blockchain is the ability to do micro payments and split that. Now obviously, it wouldn't need to be blockchain, but it was kind of thinking about how to do that without having to do other contracts and things like that.

But with the idea there, the fact that we've both come up with this idea, gives me more hope because I just haven't seen a way to do that. But what you're talking about is exactly that thing.

Essentially, it's data licensing, right?

Charlene: Exactly. Like I say, it's something we actually hope to do in the next three months as we just get into the next phase of our project.

Because we are currently in the process of turning ourselves into a commercial enterprise, but until last week, we were a research project as part of the IT University of Copenhagen working with research funding. So we are now out loose in the world, able to start monetizing and working with actual commercialization. And this is one of our main avenues that we're really keen on.

Joanna: Well, yes, because the problem I found, even just with training my brain on LAIKA, is that I don't have enough words to give it enough breadth.

So it's almost like even a writer like Stephen King or Robert Ludlum does not have the breadth of work or the number of words that would really give it a good spin in the direction of that. But a whole load of authors together would really make that possible.

The other thing is also writing in different genres. So as J.F. Penn, for example, I write in action adventure, I write in crime, I have some horror, I have some other things. So it's almost like there are ways of splitting people's IP into these different models. I think that idea has definitely got legs.

Charlene: I think we should chat about this again in a month when we're more set up and maybe we can contact you to get it rolling.

Joanna: Yeah, absolutely.

But then being devil's advocate, I know I'm a techno-optimist, you clearly are too, right. And so we're just thrilled about all this stuff, we're just giddy with excitement. But we both use Midjourney, and I'm sure you've tried Midjourney Version 4 which they brought out in the last couple of weeks. In fact, in the last week, I think, as we record this. Have you tried version four?

Charlene: Yes, I have.

Joanna: So I found with the same prompt on Version 3 on Midjourney, and Version 4, it was a step change. Like the ability of the AI to understand my prompt was just incredible. And so I feel like that is very interesting, given that you're using GPT-2. GPT-3, I did find a step change from GPT-2, and now we're hearing rumors of GPT-4, which should be coming in the next few months.

So given that, I mean, I want LAIKA and other tools to be as good as they can be, so I want it to be better and better. But again, then we come back to the issue of the training data. I mean, even you and I both using Midjourney, and the training data is very clearly in copyright. So I don't know, it's really hard.

What I don't want to do is legislate these tools out of existence.

Because you can't, they'll just go to an area where they can flourish. But how do we resolve this? I mean, obviously, we can't come up with a decision. But it's difficult, right?

Charlene: It is difficult. And as a person who's trying to set up a business, I'm very keen on us never breaking the law.

And as a writer, I'm very keen on figuring out how we can compensate writers and how we can make sure that people are able to get something out of their work.

But as a kind of humanoid on this tiny rock hurtling through space and time with only a certain amount of years to live, I just have this absolute feeling of ‘let's go, let's go, let's go. Let's keep rolling, bigger, better, faster, stronger.' 

I keep coming back to this feeling about the kind of work that is being replaced by AI or the kind of things that are going away. It kind of makes me feel something about the people who put on horseshoes in the early 1900s complaining about the car.

There's a giant world out there that we can explore and we, as the early adopters, can perhaps exploit and be part of and figure out new ways of making new kinds of money, or new ways of making a living.

But trying to just say, “this future way of doing things isn't good for our industry, so let's not let cars happen. Us horseshoe makers need to make a living,” just doesn't feel right to me.

It just feels like we need to throw ourselves into the new ways and finding out what those new ways can be.

Joanna: Absolutely. So we talked about the input there, the training data and the potential for licensing in the future. Let's talk about the output, because people who haven't used these tools worry about plagiarism.

So how does LAIKA and these other tools, how do they avoid plagiarism? Why is it not plagiarism?

Charlene: Well, I mean, it's just not plagiarism. Basically. So I'll just talk about LAIKA since that's the thing that I know.

But every time you use LAIKA, you can put in a prompt. So for example, let's say we put in a prompt of “deep down, I always knew.” Every single time you use that prompt, with any brain, you're going to get a different result. Actually three different results because we give three different inferences each time.

There is really no way other than the kind of, you know, ‘how many monkeys on a typewriter it would take to come up with Shakespeare eventually,' there isn't really going to be a situation where you and me come up with the same thing.

And so there isn't really a way in which we can be plagiarizing either each other, or anybody else. Except perhaps for like the plagiarism of the spirit of the dead who we use as our brains that we're working with.

Even still, you're never going to get like a perfect sentence that's taken from something that already exists. Every time you use a prompt, you're shifting and shaping and moving the message into a new direction that it couldn't have been in without your human touch coming in and shaping it that way.

And so at this point in time, like so what we say with LAIKA is like, whatever you make with LAIKA, it belongs to you.

Because as we currently understand copyright law, you as the human being who is creating the imagination to put in the prompt, are the only human being who can hold the copyright.

Only human beings can hold copyright.

[As we record this in late 2022.]

So we don't expect or need anybody to even say that they use LAIKA when they use it, and that's not really part of what we're interested in. So you don't have to credit LAIKA when you're working because you are still the artist. Again, it comes back to that idea of being a brush, a paintbrush, like a tool.

Joanna: Yeah, except it's a super smart paintbrush.

Charlene: Super smart paintbrush. Like every little filament of the paintbrush has its own paintbrush.

Joanna: Exactly. It will get even more powerful, and this is what I find so interesting. And I love using these tools and people listening will know about my enthusiasm too.

But you said there's no need for labeling, you believe people hold the copyright as the human. And obviously, as we record this in November 2022, the law isn't that clear on these generative tools.

There are some things, I think it's OpenAI for the GitHub copilot is going to court about copyrighted code. There are some cases, and obviously, we're going to see some cases. Again, I think this is about jurisdiction, because like you're in Copenhagen, a lot of people listening will be in the USA. And a lot of this stuff is international, right?

Let's talk about publishing, because you might not care about the output of LAIKA, that you think that belongs to the author and they don't need to label it. But how do you think readers feel about it? Or do they not care because they don't care if it's a ghostwriter?

I mean, what I've been doing because of this time we live in, I have been adding a statement of AI usage at the end of my fiction – I haven't used it for nonfiction yet at all – but saying where I use AI tools.

I have also said I use Sudowrite, I use ProWritingAid, I use Facebook advertising, I use Amazon. So I've actually included all my AI usage because people put this in one box, and I didn't think they should.

So what do you think about labeling finished work in that way?

Charlene: I mean, I think if it makes you happy, it's fine, if that makes sense. It's kind of like a content warning or a trigger warning or something like that. I suppose there might be people who aren't interested in consuming anything that has even a whiff of AI about it.

But I suppose like, I grew up with no television, so in a way, it's weird that I'm so into tech and future things because I spent my entire childhood just reading constantly. I'm a voracious reader. And if I read something and it's good, then I'm happy. And if I read something and it's bad, I put it down.

It's not really of any particular interest to me if the thing that I'm reading is written by like a man, or a woman, or a Danish person, or an American, or a completely human that has nothing else going on, or human with the help of an AI. That's kind of beside the point to me when it comes to reading something that's setting my mind on fire.

Joanna: I think partly, it's like if you sign a publishing contract, that it will say something like, “I warrant that this is my own work.” And if you submit, say, a short story to a competition, it will have a similar warrant of, “this is my own work.” And I feel like at the moment, this has not been ruled one way or the other.

So for example, I am independently published, that doesn't matter, but I do enter competitions. And when I write my stories for competitions, I do not use AI for that because I feel like then I can agree to that warrant. And I feel like if I enter a competition, and I win it, and then they find I used LAIKA or whatever, I think at the moment, the industry will consider that cheating.

So what do you think? So you said you and your partner are publishing this book with demons on the blockchain. So would you submit that to a traditional publisher?

How do you feel about the legal side of publishing and these warranties of your own work? How will that work?

Charlene: I mean, sure, we'd be happy to submit it to a traditional publisher if they'll have us. And we'll be happy to say that it's our work. And of course, we'll mention LAIKA because that's the thing we're trying to promote all the time. But I suppose there's this thing where, at least for me, it doesn't feel like cheating because it's actually really hard work to make the content into something that's shaped around a human thought.

So when we have all of our users, after they've been with us for a few weeks, we send them a questionnaire and we ask them a bunch of questions. And one of the questions is how much you agree with these statements. And one of them is, “LAIKA saves me time.”

And literally nobody thinks LAIKA saves them time, because it doesn't. But then 100% of people say, “LAIKA is fun.” And like 95% say, “LAIKA makes me feel creative.”

And that's kind of where the space is, I think. It's that you're in this playground, you're sketching, you're making something, but you have to pull it together.

If you imagine like a bunch of puppies running around in a room, like trying to corral them into the playpen is kind of what it feels like some of the times when you're trying to write with LAIKA.

You're the one who has to keep track of where the characters are, what's happening, what's going on in the plot, where things are going to go, what's the amazing theme or message or thing you're trying to get through.

And what LAIKA is giving you is some snappy phrases, some interesting suggestions and some rabbit holes that you might jump down. And then you jump down those rabbit holes, and oh my God, you've just spent half an hour writing a poem, instead of continuing with the plot that you're supposed to be on.

So in my stance, and I'm pretty sure my partner's stance as well, and in fact, the stance of our kind of copyright policy that we have at the moment, is that whatever you make with this is yours because you actually made it. You really made it. You squeezed the blood from the stone and got something beautiful out of it.

Joanna: Yeah, I mean, again, when I first started getting into this, it was 2016 when AlphaGo beat Lee Sedol at the game of Go. That's when I started getting into creativity in AI because they said that move, one of the moves, was creative, and it's changed the way the game of Go is played. And that's when I started to get into this.

So that's six years ago since we're speaking now in 2022. And I think some of my questions are probably still related to how things were a few years ago.

And we're early adopters on this, many of my audience have not even tried a text generative tool. So I feel like some of the questions that you and I might take for granted are things that the traditional publishing industry hasn't even started considering yet. Like, I really don't think that they understand it. 

One of the things I am postulating is, if you think about who has the most training data, it is a traditional publisher.

So if a particular imprint who has been specializing in science fiction, or romance or whatever, for the last 100 years, or let's say even 20 years, and they've got all that content, they could train a brain and they could then pay creatives to do it. But as you say, that's not possible right now because it's still quite a lot of work to use the tools. But where do you think it's gonna go? Because I feel like things are moving very, very fast.

Charlene: Yeah, things are moving super-fast. I mean, I do think it's gonna go into that space. I mean, I hope that we're going to be able to be part of the start of that side of the monetization of things. But I definitely think it's going to go that way.

I think some of the interesting conundrums that are gonna come out there is like when the publisher owns the work, they might start making brains like artificial intelligence brains, based on the work, and the writer might not be interested in that at all or might not want that to happen, but might not have much of a say. There could be all kinds of situations where these things are happening. Or where people are using it on the sly.

Or for example, I mean, obviously, we are people who care very much about sticking to the law and only using public domain works or works that you've written yourself, but like there's nothing stopping anybody from going and ripping the eBooks of the top 50 science fiction bestsellers and going and making a Google Collaboratory notebook and training GPT-2 on that, and then writing with it. There is nothing stopping them except for time, and patience and energy.

But there is a lot of capability out there for bad actors and good actors. And I think where we're really going to see things happen over the next few years is once the laws start getting crystallized around particular cases. God forbid that we are part of any of those.

Joanna:

This is why I want creatives, artists, writers to be involved in this discussion so that we can impact the law.

And as part of the Alliance of Independent Authors, I've been part of submissions to the World Intellectual Property Organization (WIPO), the UK Government, about this kind of thing and how AI should impact copyright.

Let's fast forward a decade, and I think all of this stuff will be much, much easier. So we're in a world of digital abundance where, like I now can create a Midjourney image, every time I want an image, I just go on Midjourney and create it, right. I don't need to buy stock images anymore. It's possible that creating fiction or books will be the same.

Therefore, I think that it's about building your personal brand, and building your relationship with your readers, and making sure your voice is strong and writing what you love to write, whether or not it's with an AI tool.

The future will be about creation, but also curation, and about building a brand where people still want to buy your books however you use AI tools.

What do you think? Clearly digital abundance is the future, right?

Charlene: Yeah, absolutely. And I definitely think that personal branding thing is key.

And I think we've already seen that happen over the last few years with publishing. Like even just with Twitter, the amount of literary Twitter over the last 10 years has just been huge. And how much it's necessary as a writer to be out there, and to have a website and to be talking all the time with people. It's part of the job now. And I think it's just gonna keep being how it is.

I suppose, I personally just don't really see – I suppose it's maybe the kind of writing that I like, or the kinds of things that I enjoy, like I really like very experimental and very weird fiction.

And I suppose that although it might seem like AI is best placed to make more of that and to make it to the way I like it, I just don't really believe that it's going to feel like that because a lot of the works that I really love feel like a conversation between me and the writer, like feel like they've touched something in me that's also in them, this resonance between souls. So I'm just not sure that I believe that AI will ever replace the kind of writing that I like, but maybe it can replace some kinds of writing.

Joanna: Such interesting times.

If people want to try LAIKA, where can they find it?

Charlene: So you can just go to www.writewithLAIKA.com, and there's a big button on the front of the page. And it's LAIKA,  L-A-I-K-A, like the space dog.

And it's basically the big button on the front page to sign up for the beta and I just add in a bunch of people every day. So usually you're spending maybe three or four days on the waitlist and then come in. But if you write in your comment, if you write something like, “I literally cannot wait,” I'll see it and let you in quicker.

Joanna: That's fantastic.

And where can people find you and everything you do online?

Charlene: So I'm Charlene Putney and you can find me at alphachar on Twitter and I'm www.alphachar.com. Always happy to get emails and I usually respond to them on Fridays. I'm a bit of an email Luddite in that way.

Joanna: Brilliant. Well, thanks so much for your time, Charlene. That was great.

Charlene: Thanks so much, Joanna. And happy writing to everyone out there.

The post Co-writing Fiction With Generative AI With Charlene Putney first appeared on The Creative Penn.