Tim: You know, you're not going to lose your job to an AI, but you might lose it to someone who is good at using AI. So, I think there's going to be room for all kinds of approaches and people using all kinds of different tools, whether it's more conventional or more AI-based. But I'm not too worried about anything going away. Because you always have to make your own opportunity and your own market as a writer.

(intro)

Welcome to the Become a Writer Today Podcast with Bryan Collins. Here, you’ll find practical advice and interviews for all kinds of writers.

Bryan: ChatGPT is fast becoming the biggest trend of 2023. But what does that mean for writers? Hi there. My name is Bryan Collins, and welcome to the Become a Writer Today Podcast. Like many, I've been using ChatGPT on and off over the past few weeks and months for various parts of the writing process. Now, I wouldn't describe myself as an advanced ChatGPT user, but I do use it for some of the administrative parts of running a blog and content websites. For example, I use it to generate iterations of headlines, to write SEO meta descriptions, to generate ideas for FAQs that I put at the bottom of articles, and even to come up with article outlines. However, I do not use ChatGPT or AI tools to write an entire article because, let's face it, the results can be mixed. I still like to insert some personality, and also insert some research into the articles that I might or writing myself, or editing, or commissioning.

AI tools can be a real-time saver. In fact, I even use an AI tool to help come up with a headline for this week's podcast episode. So, if you're listening to this, you must have found the headline at least a little bit engaging. That said, many writers and creatives from other professions, particularly artists, are worried about how AI is going to potentially replace them. To be honest, as great as these AI tools are, they're not really a direct replacement for the work of a creative anytime soon. If anything, they can be a type of writing tool that you can deploy when you want to do something that's a little bit time consuming or cumbersome, or perhaps you just want a different approach to your work.

In this week's interview, I caught up with the Canadian AI author, Tim Boucher. He describes how he's using multiple different AI tools to produce short form fiction. In the interview, he goes into elaborate detail about the tools he's using, his writing prompts, and how AI is helping him experiment. In fact, one key thing Tim said struck me. He's not using AI to produce his short stories and novels faster or more efficiently. If anything, he's using AI to experiment and explore different creative approaches. In fact, he reckons it would take the same amount of time to write one of his short stories, traditionally, as a tool for him to compile them with various AI tools.

I hope you enjoy this week's interview with Tim. Even if you're not quite ready to use AI tools or you're just tinkering around with ChatGPT, it's a good insight into how potentially ChatGPT and other AI tools can help writers in the future. I took a good, few notes from my interview with Tim, and I'm going to experiment with some of the other AI tools that he suggests which I wasn't quite as familiar with. If you do enjoy this week's interview, please consider leaving a short review in the iTunes store. You could even ask ChatGPT to write it. Or please consider sharing the show with another writer on Spotify or Overcast or wherever you're listening to this week's podcast episode.

(interview)

Bryan: My guest today is Tim Boucher. He's a hyper-realist AI artist and writer. Welcome to the show, Tim.

Tim: Hi. Thanks.

Bryan: Tim, I was fascinated to catch up with you because you're somebody who's using the latest technology, including AI, to write short books. But before we get into your process for doing just that and some of your thoughts on AI tools like ChatGPT, how did you get into writing? What's your background?

Tim: I've been working in technology for web platforms for — getting on eight years now. But I've always been a writer since I was a kid. I spent a long time blogging on different sites. But I really started getting back into writing in a more serious way when I was working as a content moderator for a fairly large web platform. That kind of work can be distressing. It can be a lot of people sort of yelling at you all the time.

Bryan: I can only imagine. From what I've read in the news, it sounds like a really difficult job.

Tim: Yeah, so, that experience propelled me back to just wanting to write and sort of exist in my own space mentally and reclaim my own creativity from all of these demands and everything that I was getting from my regular job. Doing that, that first one really unlocked a lot of doors for me, just in terms of my process and sort of building the kind of universe that I want to write in and all of that.

Bryan: I'm going to ask about the books in a moment. But before I do, you also describe yourself as a hyperrealist. Could you elaborate?

Tim: Yeah, hyperrealism is a term from post-modernism. The idea is that fact and fiction are blended, and there's sort of a blurry division between what's real and what's fantasy. I combined that in my work with the idea of the uncanny valley, where you can't always tell which parts of the work were human-generated or which part were AI-generated. It's this blurring of distinctions as an approach to storytelling.

Bryan: And you are currently writing short books with AI tools. How many short books have you written?

Tim: I just finished number 73. I started in August of last year.

Bryan: That's pretty impressive. Your short books that you're writing — correct me if I'm wrong here — they tend to average 2 to 4,000 words in length?

Tim: Yep, that's about right.

Bryan: Could you give the listeners a flavor for some of the themes and topics that you're writing about?

Tim: Sure. One of the things I like to do is to use the AI tools to write about AI. In science fiction, there's a long history of both the dystopian side and the utopian side of AI technology and how it could impact society. So, I write about that a lot. I write about this idea of a near-futuristic AI takeover of global governments and infrastructure. But I also combine that with supernatural elements, things that are fantasy, more on that side. But the way that I tell my stories, it's not really necessarily a direct, straightforward narrative.

What I do a lot of times is, I'll write books that are more like fictional encyclopedia entries. Sometimes you see this in work like the Dune series or the Foundation series, where a chapter might start with a fictional in-world encyclopedia entry, but it's usually just an excerpt. But I've gone and taken that as a primary approach instead of as a secondary add-on thing. So, I'll have, in a book, I'll explore usually one topic. I'll break it up into sections. Then each book will cross reference other books that they might drill down on to another related topic. So, if a user finds a link to another book, they will go check that out. I get a lot of people who come back, and they buy a bunch of different books just on whatever path interests them.

Bryan: That's almost what you're describing. It's much like the approach fiction writers would take to writing a series on Amazon, except you're doing this and selling direct on Gumroad.

Tim: Right. Yep.

Bryan: So, how do readers come across your books if you're selling direct?

Tim: I'm not on social media too much. There's really — mainly, Reddit is the place that I still really like for social websites. I just find that it's still fun for me. I use Medium a little bit, but I'm using it less. I'm favoring going back to my personal blog. I ran a personal blog for almost a decade and early internet times from 2003 onward. So, I'm having a really great time going back to that and being able to rediscover my own authentic voice from a blog without the distraction of the social media. And so, people are starting to find my work through that too. But I would say probably Reddit is my first place that I interact with other people. I'll pop into threads in different places and share a link or talk about something that's in a book.

Bryan: Yeah, I was looking at your blog. You're doing something on your blog that was popular a year or two ago, where people would have a page that describes some of the projects that they're up to at the moment. I was interested to read that you find WordPress a little bit weird. What's weird about it? What do you not like about it?

Tim: I mean I've been on WordPress since really the beginning, in some form or another. But it's just got too many features sometimes for just an ordinary blog. I think if you're just someone who wants to just get out there and write, a site like Medium or another one like that is going to be easier. Because you just open up a window, you write a story, and you publish.

If you're on WordPress, you have to set it up, and you have to pick a theme. Then you have to deal with plugins, which sometimes are good. Sometimes they're bad. But it's this ecosystem of third-party developers where each person adds their own contribution, and you can pick and choose. But sometimes it's hard to find the ones that are good and do the things that you want right out of the bat.

Bryan: Plus, when you install the fields, they clash with each other, break the site and are slowed right down.

Tim: Yeah, so there can be some amount of troubleshooting that goes into setting that up and maintaining that. But if you know, just to steer clear of certain things, or you have a reduced set of functionality that you want, I think it's still a really good platform. I love it because it's something that I can run on my own server. I don't have to be reliant on a company apart from my hosting company or have to worry about, is my work going to be curated? Is it going to be down ranked or whatever? I just do my thing. Then I don't have to worry about all of the platform side, which I've had to deal with too much in my line of work. So, it's a relief for me to just get back to that.

Bryan: Is the angle that you said there about the platform side, is that the reason why you're not using Substack, which many would say solves the technical issues of WordPress?

Tim: Right. I mean Substack, it seems like a great platform. I haven't used it too much. I've used it more as a reader than as a writer. I think it's probably a great way for people to reach readers and to build up a following. But it is the same thing. It's always being reliant on the decisions of a company to decide which features they're going to build or which features they're going to stop supporting, things like that. So, if I'm in direct control, I can just go and do the thing that I want when I want it or when I need it. I feel like that gives me a lot of freedom. And it takes the stress away from me too. Because as someone who has worked a lot with technology platforms, sometimes you have this clear vision of how it could be better. I'll send a suggestion to the company. They're like, "Yeah, we'll think about it." It's like, well, I'm not going to rely on you to maybe implement that thing, that if I just go in WordPress, I can usually find something that will satisfy my need more quickly than having to wait for a company to build it.

Bryan: Yeah, that's the beauty of indie publishing or blogging or doing something yourself, whatever the mechanics and tools that you use are. So, you're currently using ChatGPT. Are there other tools that you use as well, or is ChatGPT the main one?

Tim: Right now, I'm using it because GPT 4 just came out. That's really exciting. I have used a lot of verb.ai too. I haven't been using it lately, just because I've been exploring the new possibilities of GPT 4. But verb.ai, I think, is really good if you want to mix your own writing and then have something that gets completed or added by AI within the stream of what you're writing in the editor. The way that I use ChatGPT is pretty different. At this point, it's more like, I'll go in and I'll say this is the topic. This is my style. This is the format, an encyclopedia entry. Give me a list of section titles. Then I'll pick out the ones that I want. Then I'll say like, okay, use this. But add some more details. The process and the approach is really different depending on what the tool is. We can go more into that if you want.

Bryan: If you could, yeah, I'm always fascinated to hear about how writers are using these tools. Please do.

Tim: Here's one thing I'm excited about in GPT 4 too. It seems to have a better comprehension and a better approach to following instructions. It has more creative completions that it will give you. So, I found a lot of times in using the prior version, especially if you want to write something that's more fictional that has characters and story arcs and things like this, it's not great. It sorts of stumbles and it gives you really cliched things, which is part of why I sort of pivot in this direction of the more expository lore world building encyclopedia style stuff. Because it's really good for that kind of, here's something that's happening, and it just describes it. So, I really like using it for that.

In GPT 4 too, I'm really excited because the last few days, I started using it to code my own little utilities, my own little applications for customized writing tasks. I found that when I started using GPT 4, because I saw the capabilities and the quality, I started having these consistent methods. Like I said, I would have it generate a bunch of section titles for an encyclopedia entry, and then go and expand each one. I realized that it could achieve that task very well, very consistently. Then I was like, well, if I have this methodology, I could also turn this into my own little application and set up my own rules about how it could function. Then I could use the API to execute that.

But the thing is, I'm not really a programmer. I've worked in technology, but I've always worked more on the human impact side. So, I'm not a programmer. I don't have a super big background in anything beyond HTML and sort of like very basic JavaScript. But with GPT 4, I can go into ChatGPT and I can say, okay, this is what I want. I want a simple application that does this. I want to be able to enter this text and then get this kind of result. It can do it. It can take that text instruction, and it can turn it into a little application.

It takes a lot of tinkering. It takes a lot of patience. But it opens all these doors for me as a writer that I can now take my methodology of what I want to do, and I can have it become a specific tool that then calls the API and does things and returns things in the format that I want, or allows me to make a little bit more creative explorations than it might be easy or possible to do in a chat-based format. So, that's been really exciting for me. I think there's a huge opportunity for writers there as long as you are someone who has some patience, and you don't mind tinkering and going back and forth, and trying to fix errors. But it's really powerful, and it changes so much.

I've always wanted to build those kinds of tools for myself as an artist. I feel that now all of these things are possible. It's just a matter of prioritizing which ones I want to build. I'm up to my third one already. The last one I made was — if you know what an Exquisite Corpse is, it's a surrealist technique where either for writing or for drawing, each person in a game environment will write a section. Then the next person can only see the end of what they wrote.

Bryan: Oh, I'm familiar with the game, yeah.

Tim: Then the next person tries to continue the story. But then, when you unfold the whole thing, it has sort of this weird quality of all of these things. So, that's the next thing that I built for myself. It was just a test. Could I use ChatGPT to build a tool to make a creative writing exercise? So far, so good. They take time. But each one I do, it shows me better how it works. Accidentally, in the process, I'm learning a lot about how JavaScript works and how API calls work. It's not technical and annoying. It's fun. I feel that it's accelerating my knowledge and my learning in a way that I wasn't expecting. As a writer, I never thought that I would be able to achieve those goals. But now I am because of this technology.

Even above and beyond just the sheer storytelling side or the sheer text generation side, that's really enormous for me. I would like to go also eventually into those tools and be able to integrate things like image generations from Dall-E and from Stable Diffusion. Because I use those a lot in my work already. As you said, each book, it's between about 2 and 4,000 words. Then I also include between 40 and 100 images that I include scattered throughout the text, that sometimes they'll illustrate a point of the text. Sometimes they just describe a mood, or sometimes they even have another level of exposition. There might be a tone of voice in the text that when you see what's described in the pictures, you get another dimension to the whole thing.

I'm excited to be able to bring all of these tools together, both the visual side and the writing side. This is completely new, unexplored territory for me to be able to combine all of those things so easily, and have a good quality too. Sure. Sometimes the books, they're a little fuzzy on the details. Because of the AIs, they sometimes invent details, or they call these hallucinations, where sometimes from a chapter to another, it will change the name or the gender or the location of something.

Bryan: Or it could be a phantom limb in an image.

Tim: Yeah, I think if you have a conventional approach as a writer, that might become very frustrating. But I've leaned into those foibles. I want to show those things. I want to document that this is what the tools produce now. This is the way they function, and these are the artifacts that are native to this type of functioning. That's something that I think that, as a creator, you have to make peace with that, or you have to find a way to creatively use that that will fuel your goals.

Bryan: Does the application run on your computer as a type of JavaScript application, or are you running it in some other way?

Tim: You mean the ones that I'm writing with ChatGPT?

Bryan: Yeah.

Tim: I'm using just HTML and JavaScript. Then it has some API calls and a JavaScript file. So, it runs on my computer. It's the kind of thing that I can easily share with friends, so they can run without any special setup too if they want to test it, or if they want to make their own version of it. Eventually, the ones that I make that are less buggy and that are a little bit better quality, I'll probably put on GitHub so that other people can try them too and plug in their own API keys, and see what they can come up with.

Bryan: Fantastic. Last week, I was reading an interview with Claire Silver, who's an AI artist. She's credited some of her work as NFTs. So, I was reading about how she creates her AI art. Sometimes she'll get imagery faded into one of the tools, generate an output and then take that output and put it into a different tool. Is that something you've experimented with?

Tim: I have actually. One of the things that I started to do with another tool that I built the other day was, the tool would generate chunky bits of text that were on their own a bit incoherent. But I would take those outputs from the app that I built. Then I would bring them back into ChatGPT using GPT 4 as the model. Then I would say like, okay, now make this into something that actually makes sense but using the same elements. It did a great job. It just took this thing that was like hazy, cloudy in it. It turned into something that actually you could read and make sense.

I've done that too with bouncing back and forth between tools from completely different providers, like taking something from verb.ai and bringing that into character.ai, which is another one that I like. You can set up fictional dialogues or fictional bots, and then have a dialogue in character with that bot. So, I've done that too, like taking the dialogues from character and then bringing them into another tool, and then having completions that are generated by a third tool. I think it's fun and it's interesting. You get results that you won't get otherwise. But sometimes it's time consuming to pop back and forth between all of those things when you just want to get to the finish line on a specific project. So, I think there's room for both with the exploration and the combination. Then sometimes you just want one thing to get the job done.

Bryan: How do readers react when they find that you're writing these short stories and snapshots with AI?

Tim: I've had a range of reactions. I had a couple major news outlets that they wanted to do fact checks, because I published some sets of images that I made using DALL-E that they were supposedly depicting the rediscovery of a lost civilization in Antarctica. Those images, they were labeled in the post that this was generated by AI and all of that. But other people took them, and they re-contextualized it either in Facebook, or YouTube, or TikTok, or Twitter, or whatever. They took out the watermarks in the image that show that it's from DALL-E. They didn't ever reference my original post.

A lot of people were looking at those images and imagining or hoping that they were real. That's why Reuters and France 24, they contacted me to do fact checks because they followed the trail back to me of the images. So, I've seen a range of different reactions, especially to those things. Because people see them in the wild, and they don't know necessarily that it's connected to a book. And so, some people will be like, "Oh, this is definitely real." Other people will come in and say, "It's fake, and here's why." They'll point out all of the things. I like that. I want there to be this conversation about the reality and the unreality of the whole thing. That's the niche that I'm after.

In my work, I do a combination of different types of labeling. On Gumroad, I label myself as an AI publisher. My brand or my imprint is called Lost Books. Then within the books themselves, there's a disclaimer that says some aspects of it may be created by AI. Also on my website, I write about the books. I write about the process and about building these different tools and about using AI and my reactions to the industry and how all those things are going.

I think there's a lot of different pathways that people are coming in to the books. I think some people are buying them knowing full well that it's AI, and that it's just like a fun sort of art book. Then maybe there are other people who are coming in from a different pathway of finding it. They might have the idea that maybe some of this is real. I don't necessarily, as the artist, want to tell them which is which. That's part of the intrigue for me. It's mixing those up and having people make their own assessments about what's happening here.

Bryan: It's pretty clear when I look at Lost Books that AI is helping you to create these stories. If I were to buy one of these, I certainly wouldn't feel misled. Typically, anyone who borrows AI worries that it's going to replace the work of writers. But it sounds like your process is quite creative. You're coding. You're combining writing prompts. You're finding approaches to working with different tools. Do you think these tools are going to become a more core part of the writing process for fiction writers over the next few years?

Tim: Yeah, I mean, absolutely. I don't really adhere to this idea that people are going to lose their jobs over this. But I have seen one take on that, that I like a little bit better. That says something like, you're not going to lose your job to an AI, but you might lose it to someone who is good at using AI. So, I think there's going to be room for all kinds of approaches and people using all kinds of different tools, whether it's more conventional or more AI based. But I'm not too worried about anything going away. Because you always have to make your own opportunity and your own market as a writer.

Bryan: If somebody's listening to this, and they're there thinking, "I'm going to start experimenting with AI," is ChatGPT the tool that you'd recommend they use first, it's the one that gathers the most news headlines?

Tim: I think it is good because it gives you an idea of the scope and the depth of what AI is capable of. It's the most versatile in terms of, like, you can just give it a plain text description of what you want. It will give you something that sort of matches it. Then you can go back and forth and fine tune it. But I really do verb.ai also, just because you can go in with a paragraph and then use a command continue, and it will just write the next thing that it thinks is going to be there. Then you can also do a describe command.

So, if your characters see an object, you can say like describe this object. You give some parameters for the description, and it will do that too. It's really simple. It's really basic, but it's starting to have some powerful functionality too. You can use the outline mode to describe characters and to describe what happens in each scene. I'm bad at that kind of thing. I'm not a good writer in terms of planning out plots and stuff ahead of time. But I know some people really like that kind of outlining. I think those aspects of it are going to become better over time. I think there are still a lot of tricky technical challenges in making that work.

I would recommend those too. Another one that's free that I use a lot is textsynth.com. There's a playground tool on there where you can pick between a few different open-source language models like GPT-J and NeoX and some other ones. Between those three sites, I think most people can get what they want. There are a bunch of other ones not to. I've tried a lot. But those are my favorite ones.

Bryan: Yeah, I'm looking at TextSynth here. You can use this without signing up on accounts. You can just paste in some sample text, and it'll automatically produce an output. What about the commercial tools that I read recently that Jarvis AI may have a billion-dollar valuation, which seemed incredible.

Tim: It sounds like a lot of money, yeah. I have seen some of those tools. I've used some of them a little bit.

Bryan: Or jasper.ai, I should say, I think they rebranded.

Tim: Yeah, Jasper. I'm less interested in those tools, because they are more aimed for marketing and search engine optimization. I know you can use them for other things. But I think that's their biggest market. That's probably why they had such a high valuation. It's because everybody wants to do marketing copy. It's a kind of writing that it takes a special type of skill, and not everyone has it. So, it brings things into reach that were not in reach otherwise. I think there's a utility for it. I don't use it myself, just because I'm more interested in the experimental, hyper creative side of writing and image creation and all that stuff.

Bryan: When it comes to the image tools, do you have a preference for which software or tools you'd like to use?

Tim: Yeah, I use just regular DALL-E. But also, I've been using a lot a tool called playgroundai.com. It has DALL-E that you can use, and you pay credits for it. But it also has a free daily limit for making stable diffusion images, which I forget what it is. But it's something like 1,000 images per day for free, which has an enormous amount of images per day. I've used both their free version and the paid version. I think most people would probably be fine with the free version, if they just want to see what it's like and how it works if you haven't tried it before. I've tried Dream Studio too, which is the one that's like the official one from Stability AI who created Stable Diffusion or helped create it. I think actually, Playground AI has a better interface. It's more friendly, and it's free for so much use. So, that's primarily what I use.

Bryan: What was the playground URL that you described?

Tim: Playgroundai.com

Bryan: Playgroundai.com. Oh, great. I'll be sure and check it out. How long does it take you to produce one of your books using all those tools and systems?

Tim: Yeah, when going at my fastest, I can usually get a new one out every three days or every four days or something. But I think in terms of hours spent, it's probably between 6 and 10 hours per book.

Bryan: Which is interesting. Because it would take, I would say, a professional writer about that long to produce something traditional written without the help of AI as well.

Tim: Yeah, and that's something I thought of before.

Bryan: It's not like you press a button, then it spits out a story.

Tim: Right. It's like, okay, for a 2,000-word piece, am I spending the same amount of time that I would be doing it manually? Sometimes the answer is yes. Sometimes the answer is that it allows me to greatly increase my word count. But sometimes it's not that, that I get from it. Sometimes it's sort of the direction or the content or the quality, though, that I'll get from the tools. But also, that includes generating images. A lot of times, I'll generate close to double the quantity of images that I'll end up using. Then I'll go and I'll sort of curate the best ones.

Then I have some image processing that I'll do for cover art and for previews that I use in Gumroad to show what kind of art is in the book, and what are the major themes and stuff. Then just, I use Vellum as an ebook app. I'll combine all the final results there and generate that as an EPUB or as a MOBI file. Then I put it up on Gumroad. And basically, I'm finished.

Bryan: Readers find the books then organically via Gumroad and also via Reddit, from what you were saying earlier on. Interesting. It's a great approach. Tim, if people want to read your books or learn more about your thoughts about AI for writers, where should they go?

Tim: Sure. They can go to Gumroad. And if they type in lostbooks.gumroad.com, they'll find the work. You can also go to my publisher website, which is lostbooks.ca. That will give you a link to the Gumroad and then a link to some other books and some press that I've had. It will also link to my personal blog, which is timboucher.ca. That's spelled timboucher.ca. That's where I'll write about the technology and the art and the books themselves and the process and everything.

Bryan: I'll be sure to include the links in the show notes. Thank you for your time, Tim.

Tim: Yeah. Great. Thank you.

(outro)

Bryan: I hope you enjoyed this week’s episode. If you did, please consider leaving a short review on the iTunes Store or sharing the show on Spotify, Stitcher, or wherever you’re listening. More reviews, more ratings, and more shares will help more people find the Become a Writer Today Podcast. And did you know, for just a couple of dollars a month, you could become a Patreon for the show? Visit patreon.com/becomeawritertoday or look for the Support button in the show notes. Your support will help me record, produce, and publish more episodes each month. And if you become a Patreon, I’ll give you my writing books, discounts on writing software and on my writing courses.