#273 – The Future of AI Writing with Shane Neeley

Author Shane Neeley joins us on the podcast this week to talk about his experience using AI and machine learning to help him write his books. Shane also discusses his strategy of self-publishing, the future of writing, and whether robots are going to take over the world.

Author Shane Neeley joins us on the podcast this week to talk about his experience using AI and machine learning to help him write his books. Shane also discusses his strategy of self-publishing, the future of writing, and whether robots are going to take over the world.

• Shane tells us about his career change from a biologist to an AI Software engineer and data scientist and talks about natural language process systems; feeding information to a software so it learns the way you write.
• He explains how AI helped get rid of his writer’s block and the positive and negatives of having an AI autofill plotlines and paragraphs, including a robot’s sense of humor at the end of each chapter.
• He discusses his book ‘The Stone Age Code’ and how he used AI Art to produce the cover as well.
• Shane talks about the self-publishing aspect and why he decided to go that route. He also discusses marketing strategies that he used for his books, including advertisements and social media.
• He talks about using AI for audiobooks and whether that could be a future project to try out. He discusses the difference between AI and humans voices, and if you’d really be able to tell the difference.
• Shane talks about the tools he used for writing and what he plans in the future using AI.

Useful Links

Shane’s Website
Stone Age Code
Based on a True Story
Don’t Drop the Mic

Shane lives in Oregon, writes code and non-code (books / blog). A former laboratory scientist turned programmer, he helped to author 11 scientific papers and build a cancer treatment software company. Having traded in his lab coat for a laptop, now using machine learning and artificial intelligence for everything from bioinformatics to art. He is particularly interested in AI for language generation, digital art and other means of computational creativity.

Shane shares a transformative message about technology and biology through comedic writing and speaking. Always trying to entertain the reader, communicating advanced topics with a spectrum of joking.

Episode Transcript
Transcription provided by Speechpad

Joni: Hey, writers. You’re listening to the “Kobo Writing Life” podcast where we bring you insights and inspiration for growing your self-publishing business. We’re your hosts. I’m Joni, author engagement specialist at Kobo Writing Life.

Rachel: I’m Rachel, author engagement coordinator at Kobo Writing Life.

Joni: On today’s episode of the podcast, we’re talking to Shane Neeley, author of the “Stone Age Code: From Monkey Business to AI,” and also the book, “AI Art and Poetry.”

Rachel: We had a really interesting conversation with Shane about both of his books and how he uses AI and machine learning to help him write his books, create the art in his books. And we talked a lot about where AI is going to fit in, in publishing now and in the future. And don’t worry, not replacing humans quite yet. But it was a really interesting conversation, and we hope you enjoy.

Joni: We’re here today with Shane Neeley, the author of the “Stone Age Code.” Thank you so much for joining us.

Shane: Thank you, Joni, thank you, Rachel. I’m really glad to be here.

Joni: Can you start by telling our listeners a little bit about yourself and about your book?

Shane: Yeah. So, I started my career in laboratories, as a biologist, doing the typical dissections and pulling out fused rabbit spines, and creating new viruses to inject monkeys with, and all of that. And after a few years of that, I thought, “Yuck, I don’t like this lab work anymore.” So, I started to pick up some programming skills. And I was able to teach myself how to code and impress my boss enough that he let me stop doing all that, that mouse torture and all of that, and start just programming instead. And so that was about a decade ago. And, you know, ever since, I’ve become a software engineer and data scientist. Right now, I’m at a company that does cancer research, search engines. And yeah, during the pandemic, I use some of these natural language processing skills that I had gained in the search engine to help me write a book with AI, write two books with AI. So, yeah, that’s what I’ve been doing and became self-published.

Rachel: Can you kind of explain to our listeners what, was it, natural language process systems are?

Shane: Yeah. So, in this search engine world where you’re ranking documents, I use these techniques to determine the quality of documents, where the computer itself is actually reading, and, you know, determining the quality and the class of documents. And the algorithms for that in recent years have gotten really good and intelligent. And along the same lines, the people who invented these algorithms, a lot of them are in places like Google, or OpenAI, which is an Elon Musk spin-off of an open-source AI company. And they’ve released algorithms that not just read but they can write as well. And over the last couple of years, Iet’s say since 2019, those started getting really good, and nobody was really quite expecting algorithms to write as well as they do.

And so it started, you know, making headlines because you could write a ton of fake news, and you can just title a blog post or an article, you know, whatever, “Donald Trump talks with China.” And then they have just a bunch of fake news written by a robot that sounds legitimate, but it’s totally contrived. And I started paying attention to those. And then when the lockdown hit, I had some time to actually play around with them. And I found that, you know, it doesn’t have to be all fake news, it could actually be funny, and it can actually help me do something I’ve always wanted to do, which is write a book.

And since these have come out and improved, I’d say, writer’s block is dead at this point because it’s so much more about just…you can get so much information out of it, and it can fill out your plotline, it could write the twist, it can describe characters, and it’s much more about, it’s more of a readers block where it can just keep writing and writing and you just have to decide what is best to incorporate because it can write a lot of junk, but it does help you get through. And there’s new tools out there to do it. Me being an engineer, I, you know, kind of rolled my own and did it myself. So, the one I used for both of my books this year was called GPT-2, which is an open-source algorithm that you can train yourself.

So, I had previously spent about a decade, you know, when I wake up I start journaling a little bit and just to kind of get my mind going. And so I had about a decade’s worth of just little writings and journaling. And so I had a corpus to train the algorithm so that it could sound like me and become a little clone of me when it writes. It writes as if it sounds like me after training it. So, that’s what I did with one of my books, and it turned out fun. I don’t think it made the process any easier because writing, and self-publishing, and marketing, and, you know, forming the company, and all of that was still a ton of work, but it gave me an interesting angle to become an author. So, I’m really happy with that.

Joni: So, you say it didn’t make the process easier, but it also eliminated some of the writer’s block for you. Can you explain that a little more?

Shane: Sure. When coming upon…when you open up your document, your manuscript, and you see the last thing you’ve written, and you don’t know what to write next, really, you can copy that last paragraph and paste it into one of these algorithms. There’s ones you can use commercially, one’s called Sudowrite, there’s a…Luther.AI came out with one you can use. There’s a bunch out there, and you paste it in, and it kind of autocompletes the next few paragraphs for you. And they may be junk, so you may have to decide that makes no sense, but try again, try again, try again. And maybe you get, “Oh, that’s an interesting detail they added to the character, I can follow through with that.”

For me, I took the approach of, I was the serious author, and I was writing about how to learn how to code, how to learn how to do machine learning, why you can do it because of your evolutionary history. It’s a strange book in that way, but we can get to that. But, for me, I put the robot writings into their own excerpts so that I can be the serious author and then the robot could say something ridiculous that I can kind of poke fun at. I did it for comedy, but a lot of people are doing it for either actually, you know, coming up with new ideas in the plot and new details. But, yeah, I just wrote my non-fiction and then had the robot write something crazy, just kind of make the book a little bit more fun.

Because that’s one of the aspects of these AI methods is that they write totally strange off-the-wall things because they have no sense of truth. They don’t care about truth. They kind of sound like they’re a….at least the one I use, it kind of sounded like…it was like, imagine you had like a psychology paper to write, and you had to get it done and you procrastinated to the end, and you just, you know, drank a Red Bull the night before and typed out a bunch of garbage and submitted it. That’s what I kind of think of these things because they don’t quite make sense, but it kind of sounds okay.

Joni: They are funny, though.

Shane: Yeah.

Joni: So, one of the things that’s really cool is the way that you can make it sound like you are, like, whoever it is, like you can kind of…and that’s by training it, by feeding it your corpus of use language, right?

Shane: Yeah.

Joni: Does it improve? Like, do you constantly feed it a new language? Or do you sort of set it up with a ton of information and then let it do its thing?

Shane: Right. Yeah. So, they start being trained on…you know, they say, like, the whole internet, it’s trained on the whole internet. So, the model starts with this big brain and understanding of language and writing. And it can write just like that. It’s seen a lot of things, it can just write. But to make it sound like you, you have to fine-tune it on your own writing. So, you may be an author with 10 books, and you basically would put all of those 10 books into a single file and submit it to the process for training so that it sounds like your books and understands a little bit more about your characters and your style.

For me, yeah, I did periods of retraining where, as I was writing the book or writing new blog posts, I put it back into the training data and ran it again and see if it sounds a little better. And probably the 10 retraining throughout the process, which, from my perspective, it’s kind of expensive because you need really advanced computer hardware in order to do the training. And so that’s what made mine different is I was training myself versus there’s a lot of commercial ones out there that you just use but you don’t do the training because they haven’t set up the methods to do that. And it gets expensive because of the hardware you need.

Rachel: Could you tell our listeners a bit about what your book “The Stone Age Code” is actually about, this book that we’re talking about, theoretically right now?

Shane: Sure. I always wanted to write a book on human origins, because it’s always been interesting to me of, like, why we act the way we do, you know, who are we? Our previous genetic diversity fascinates me compared to our current complete lack of diversity, you know, no matter what race you are. We all come from a very small population of humans that came out of Africa 60,000 years ago. And some groups are a little more diverse than others, but I was always fascinated by the fact that the Earth contained Neanderthals and Denisovans and, you know, possibly 20 completely different humans from you and I before with large brains, and they spent a lot of time making tools and technology, as well. We happen to survive, could be luck, could be a little bit better skills, social skills, or technology skills, but I just always wanted to write about that.

But I’m not a paleoanthropologist so I couldn’t do a book completely about that and have any authority. I’m an AI engineer, and so I took the angle of using those analogies of, you know, these early humans making technology. And nowadays, people like me are still doing that, where instead of sitting down to chip the perfect stone in order to stab a rhino, or skin a deer, or whatever, we’re still sitting down and chipping that perfect piece of technology, which I think allows me to sit down and write code every day or write a book, this sort of strange obsession, where you sit down and just try to improve something over and over, come through from who we are.

And so I was able to write that analogy between who we are and how that technology leads into our most advanced technology today, which is probably AI. And, yeah, and there’s another little bit of a subplot in there about art and where art comes from. And the book, “Stone Age Code” it’s full of…it’s not full of, it has about 11 images of AI-generated art. And once I started finding these AI-generated art methods, I got completely distracted because it was so fun to do as well. So, I did the AI writing, and then I did the AI art. You know, I’m not an artist, but I felt like I had a good aesthetic sense in order to see what looks good.

So, that helped me produce the cover, and it also completely distracted me for about four months during the writing of this book, where I just produced a coffee table art book, which was my…it was a second book I started on but the first I put out called “AI Art-Poetry,” where the AI responds to an AI-generated image, and with a poem about it. That’s the other book I put out. And so, yeah. My cover, it has a rock in just an early stone tool that is styled to look with the colors of the Spotify user interface because I thought that looked cool. And on the back is a monkey on a rock, where the rock itself is styled to look like a Linux terminal. On the back of this book is an ancient hand painting from a cave, but it’s also styled to look like a Linux terminal. And so the book is just full of…you know, I’ll take a gorilla playing with his child and style it to look like Cat 6 internet cable. So, the gorilla looks like it’s made out of these cables, and that’s the AI-generated art called style transfer as well.

Joni: That’s very cool. I kind of get my head around the idea of feeding an AI machine a lot of text and having it learn that way. How does it work with artwork? Is it the same concept that you’re giving it a lot of different pieces of art?

Shane: Yeah. It’s similar to the writing one where the writing one is fed on the whole internet and it learns how to write. These other models, they’re fed on millions of images and they learn about, you know, what eyes look like, they learn about what lines look like, what all these other things in photography, you know, what sunlight should look like against this. And so they gain an understanding of imagery. And when you use it, it’s called transferred learning, so it’s already been taught, and then you’re transferring it your own. And with style transfer, it’s an algorithm that takes two images and it tries to…one is the content image and one is a style image, and it just tries to apply the style of one on to the content of the other. And a lot of the early examples were like, “I want to make a Monet look like a Van Gogh.” And it’s interesting, like, if Van Gogh were to paint Monet’s painting, how his style would have been applied to it. And so, yeah, I had fun with that aspect, as well, and takes a lot of computer hardware and money in order to do these.

Joni: It’s very cool. It’s really creative, so it needs your input, right? Like, you need to say, “Okay, I would like these two things to be blended and then the machine regurgitates something based on what you’ve put in?”

Shane: Yeah.

Joni: Okay.

Shane: The first time it does, it may not look good. And so you sometimes have to try like hundreds of times, then you get the one that you really like. So, I knew my book was going to be printed, and so I needed black and white images. So, I took a picture of a polar bear, and I believe this one is styled to look like sheet music. So, it was a polar bear, it was styled to look like sheet music. It’s not made of, like, the notes or anything, but it gave it that black and white style of sheet music that I could use for printing.

Joni: Do you think this is something that will become used more in publishing, particularly with indie publishing? Do you think that there’s a way to make this or is it, sort of, inaccessible for most people right now?

Shane:The way I’ve done them is fairly inaccessible for most people. And people have said, “Why don’t you make a program for authors to help them create their cover and styles and help them train their own algorithm to do their own writing?” And I could, but I have a day job and I think somebody else is going to do it before me. And you’ve probably heard of Joanna Penn, and I was on her podcast as well. And she gets a big list together of all these AI aspects that you can use in self-publishing. She has a website that she curates with, like, dozens and dozens of examples. So, there’s a lot of it that’s very accessible for the average author that you can find there. And for me, it was just following my own curiosity and being a programmer and doing that, but it’s definitely gonna be available to more people.

Joni: Even if you don’t use it for your final cover, I like the idea just for brainstorming to be like, “Well, what does this, how does this work, what does this look like,” and testing through, like, putting what’s in your head, I guess onto the screen and testing different things, I think it would be really cool.

Shane: Yeah. And I use it for, kind of, marketing as well, because a lot of the images that I did not include, I just put them on Instagram and #AI Art and Style Transfer and get some exposure there. So, it makes a digital artist out of somebody who doesn’t really know anything and doesn’t know how to make digital art. So, in some ways, it’s cheating because I used photos, I didn’t draw the images, but they look drawn. And it was my aesthetic sense that decided, out of the hundreds here, what was the best-looking ones.

Rachel: I just want to kind of touchback on the content of your book, “Stone Age Code.” You said that you are not a paleoanthropologist, but there is a lot of paleoanthropology content in your book. What was the research process like to, kind of, make the comparison between AI and human evolution?

Shane: Sure. Yeah. A lot of my curiosity started many years back reading books such as…there’s one called “The Mating Mind,” it’s about sexual selection in prehistory. What created art in the first place is somebody doing a cave painting to attract the opposite sex, and those who are better at it were able to mate more successfully, which…you know, we do a weird thing called art. And so this was a whole hypothesis of where did that come from? There’s another one called “Man’s Emerging Mind.” There’s a book called “Sapiens” is a very popular one out there. So, just having read those years before, I wanted to learn the subject, so I just started writing about it and quickly find out like, I don’t know anything. There’s so much change in the field, just them doing the Neanderthal DNA 10 years ago and the Denisovan DNA less than 10 years ago. They’re finding out so much new stuff. Even the pros feel like they don’t know much at this point. So, it’s just a very evolving and fast aspect. And “Kindred” was another book that really inspired me, it was a recent book, very detailed about Neanderthals, Rebecca Wragge Sykes.

And when I started writing about this stuff and trying to tie the analogies together of how we’re evolving towards AI, you know, out of the Stone Age into the Code Age, I approached an editor that I found on Thumbtack, actually, and I found two, and I was like, “Oh, this one’s an English professor, he seems really good.” And then this other one was an English professor, he also seems good. And then he tagged his email with, “Oh, by the way, I’m also a great-great-grandson of Charles Darwin.” And so I thought, “Whoa, okay, this kind of, you know, giving me goosebumps.” I was like, “I’m writing an evolution book, and I found a relative of Charles Darwin.” His name’s Adam Cornford. And he became my developmental editor at the time as I approached him early and the book was less than half-finished, and just showed him my Google document, and he read it and had so many unique perspectives to add to it, as well, you know, knowing his great-great-grandfather’s work, obviously, and he’s also just interested in various societal aspects, and really helped me out a lot and find the direction of the book.

And then I approached another mentor during the book, Emiliano Bruner, who is a real paleoanthropologist who studies the brains of early humans and the skulls of early humans, and he loved the book as well. He’s a guy in Spain. He really loved what I was writing and he helped form the direction for that as well. And so it came together because you have previous interests in this stuff, and then I found the right people that were willing to go into my manuscript in Google Docs and help me formulate the thing.

Rachel: I think that’s really cool. And as somebody whose science kind of topped out at grade 11 biology and whose coding experience was like Live Journal in the early 2000s, the comparison between the evolution of humanity to the evolution of AI was really helpful in helping me grasp the content of your book. Did you run into any difficulties making everything so clear so laypeople like myself and Joni could understand the content of your book?

Shane: Yeah. That was actually really hard because it was my first book, for one, and so I wasn’t exactly sure how to write to an audience. It was…people’s first books are often a passion project, like what they really want to write, what I had in me for the previous 30 years kind of comes out in this first book. And maybe my next one will be more directed to a specific audience. But I just distilled the research, both in machine learning and paleoanthropology, into what I thought was…what was accessible. But I didn’t remove the jargon, I wanted to keep in all of the jargon around deep learning. And just in case you were going down that path towards learning how to code or learning how to do machine learning, at least you would have heard the jargon once from me with the analogies, how I compare it to monkeys, or, you know, a big machine with millions of dials on it that monkeys are tuning.

I compare that to the algorithm and you kind of smack the monkeys to do it faster, that’s, you know, compared to the learning rate of the algorithm. And these jargon and deep learning, I just wanted to keep it in there in case if that was your path was to learn how to code in the future, or maybe you’re an engineer who wants to get into AI. But, yeah, the audience, that was a hard part, finding out who to write this to, but luckily, those couple of mentors liked it and they were more scientifically based. So, I think it’s probably a little bit less general readership, a little bit more science and code-based readership.

Joni: Maybe. But as Rachel said, I thought you really nailed making it accessible and understandable to non-scientists without seeming like you were talking down to us, as readers. So, I liked that.

Shane: Yeah, thank you. That’s definitely where the robot really helped because at the end of each chapter, I’d have it do something just funny and I respond to it. And then it’s also a bit memoir, so that doesn’t take any advanced readership where it’s just me talking about my career in this and that. So, yeah.

Joni: And the robot adds levity, for sure. It’s fun. And I’m interested also, did you always plan to publish independently? Was that part of the plan with the first book?

Shane: Yeah. Well, I didn’t know how any of this works. Once I started putting words on paper, I may have early on had the idea of, “Okay, you write the manuscript and you submit it.” But I quickly found Vicki Fraser and Joanna Penn, and also my girlfriend’s friend is a self-published author, Meg Muldoon, of a whole bunch of Cozy Mysteries. And that’s her whole thing, that’s her full-time job, is self-publishing all these books. And so I learned that it is possible to do it and then the community was just amazing. I read a bunch of books and podcasts, especially, and to slowly put these pieces together. And eventually, I formed the publishing company to put these two books out just to organize my finances and figured out, you know, I could do this all myself. It definitely took a lot of work organizing all the contractors, all the editors, copywriters, proofreader, cover design, interior design, all of those people that I…even freelancers that I found. So, it was more work than I thought it would be, but I just got kind of obsessed during the pandemic and figured it out. And so now I’m confident I could put out any book myself.

Rachel: That’s awesome. And then you have quite a bit on your blog about what it was like for you to actually sit down and write this book because you planned to do it for a long time. And so you’ve got a lot of good tips. Was there anything in particular that really helped you to actually get it from concept to page?

Shane: Yeah. I think just getting that word count up gave me confidence that I can do it. Just looking up, how long is a book? Okay, a non-fiction book doesn’t have to be 200,000 words, a non-fiction book can be less than 50,000 or 30,000 even. And so once I saw that word count creeping up, I knew it could be more than just an article, and that gave me confidence. And then the other thing that really spurred my confidence to keep going was just the idea of having a book and getting…early on, you know, when I maybe only had 10,000 words, I contracted the cover designers. And it was just once I had them come out with a cover, and the 3D mock-up images of the book that looked really thick, I thought, “Dang, I’ve got to keep writing this thing now because it’s a mock-up image that’s got a lot more pages that I currently have.” And so that just spurred me on to keep going.

But, you know, putting these two books out, it wasn’t easy, it definitely took a lot of…there’s just so many boxes to check. And then to finally get to the marketing, which was also crazy. But I think, just, you know, I would just say, just keep going, and whatever keeps inspiring you, like, I found those mentors or beta readers that really helped me keep going, and the developmental editor helped me keep going as well.

Rachel: And you mentioned that your marketing was crazy. What was your marketing strategy for your book?

Shane: At the company I work at, I helped manage a lot of the marketing. Well, we do things like Google ads and paid ads, and we also do a lot of social media at the company. And so I thought I had a bit of an idea of how to market. And I saved a bunch of money for it, for the launch, and took this shotgun approach of paid advertising on every platform, and paying freelancers to help me make the ads and put them out on, you know, like five different platforms and just launch the book, and to share a lot on social media. And it took a lot of money, and I’m still trying to recover the expenses from that because it did not hit the way I thought it would, and probably because I wasn’t so focused on maybe just choosing one way of doing it. So, the launch and marketing, certainly, that was intense.

And at this point, I found out how to slowly sell books more consistently and not spend a ton of money on it. So, I’m happy where that’s at now. But, yeah, I think the self-publishing community as well also puts a ton of emphasis on the launch. And getting your launch team, you know, I put a launch team with, like, 30 people to share on their social feeds and this and that. But, you know, being a first-time author, I feel like it was very hard to really move the needle despite the fact of having a lot of advertising and this and that. So, now, you know, I’m just taking it slow, and thank you for allowing me to do this podcast as well. And I think, you know, this is probably a good strategy. It’s really just connecting with people and doing stuff more organic, less paid advertising. That’s my plan right now.

Joni: I think it’s probably also you almost made it more difficult because you have one book that works great across all platforms, but your art and poetry book really is a print book, which is a lot more challenging when you’re doing self-publishing. So, how did your strategies for selling differ between those? Because I feel like the indie world was quite focused on digital.

Shane: Yeah, I was excited once I could get “Stone Age Code” out and put it at a decent price, you know, put it at $6.99. I could keep lowering it and still make profit. But the poetry book, it’s 200 pages of full color, and it’s a 10 by 10, and it costs me a lot to make it, so I have to sell it for like 30 bucks. It’s more of a coffee table book. And so, yeah, and then you don’t get the aspect of, you know, easy click download like you do for the digital. So, digital sales, definitely is good, and I definitely want to play around with the price and…You know, I went wide with both books, wide print, and then wide digital, and there’s so many more options to do wide digital, and the international aspect’s always excited me. It’s just like, I’ve always wanted to get the book out into just a diversity of hands, as well. So that’s why, you know…I get most satisfaction out of, you know, somebody in a completely different country buying and reading the book, or some library picking it up that, like, you know, just being somebody who frequents the library, that excites me. You know, even though that’s not a big sales aspect, but just some library picking it up, whether digital or print.

Joni: Yeah, that’s awesome. You’ve got a slightly niche audience, which is almost a good thing when there isn’t that much in the market that’s like your book. And I think with digital art, especially with…I feel like there’s a lot of attention on digital art, and NFTs, and that kind of thing as well right now, which makes it sort of a good time to be stepping into the space, so.

Shane: At the point of the launch was a very peak in NFT value. And right before the book was published, I made every image in it one of one NFT…

Joni: Oh, yeah? Cool.

Shane: …and put them out there thinking, “Okay, this book’s gonna make me famous, and then people are gonna want…I’m gonna own a piece of this book, and nobody else is gonna have it because I own the cover image, I own this and that.” And so I put those out there, and it really helped me, it get a lot of impressions. You know, it was an interesting story to take, okay, this is a book, but it’s also digital art that you can own. And that was fun, and that’s…yeah, I continue doing that.

Joni: Have you looked into audio AI at all? Because I know that that’s something that we talk about quite a lot at Kobo, the kind of introduction of audio and whether or not it’s convincing as a human narrator.

Shane: Yeah, I’m very interested in it because, well, for one, I want to make the audiobook and I plan on putting the audiobook out this summer. But getting a studio and editing, and all of that for the audiobook together is a lot of work and expense as well. And so if the AI-generated audio…I know there’s ones that can be trained to sound like you, but I think you would need a lot of your own voice put into it first. And so, yeah, maybe podcasters and this and that can do that, or you just have somebody generic read it for you, or people can decide what voice they want to hear your book in. I’m not sure if it would work for [inaudible 00:32:30] because I’m a big sarcasm kind of guy. And I’m not sure the jokes between me and the robot would really come out in audio. So, I’d have to try that out.

Joni: Yeah, it’s the kind of book you could narrate yourself because it’s memoir-ish.

Rachel: I have an AI question for you. Do you think that AI will get to the point where it can convincingly write a book to the point where, like, human readers won’t know that it was created by an AI?

Shane: I do. I think it’ll be a bad author at this point. I had a friend who’s using a different method and we tried to, you know, quickly write a book, it may be more like a blog post, but he was just like, “Okay, let’s write about horses and put it into it 10 ways you can buy a horse.” And the AI was quite convincing as to what the 10 methods you would need to buy a horse. I don’t know how it knows this stuff. I don’t know anything about horses, but it was like you got to get a stable, you’ve got to get a veterinarian locked in, you’ve got to have the land for it. And it sounded very convincing, at least in this article format. And so I could see it really completing a full chapter. Basically, you decide the headings of the chapter, maybe you write the first paragraph of the chapter, and it completes it.

But obviously, you know, the difference in quality of books is just immense. And right now, there’s so many books out there that you just don’t want to read, and I think AI is just going to make more and more books that you don’t want to read. And I think people are going to do it. People are already putting out books in the stores that are 100% written by AI but they changed their pen name. They didn’t want it under themselves. They put it out there just to see if it would sell them but they change their pen name. And I don’t think they’re really any good compared to, you know, the great books out there. So, I think it’s gonna come down to…there can be a flood and flood of AI-written books because it can write billions, it does write billions and billions of words per day, at this point, lots of fake news, lots of books. But these algorithms are maybe, at this point, writing more than humans on Earth are writing because they can just write so fast.

And so there’s going to be a lot of books, but I think it comes down to using AI recommendation systems in order to choose books that you would like. So, the Kobo readers, like, they have their history of books that they like, you would need a recommendation system, which you may have, to recommend books that are similar to what you liked before this quality and type of book. But it may need to go deeper where the AI reads the book itself and then decides, not just based on metadata, or alsobots, or this thing. So, I think, yeah, the solution to AI writing too much will be AI recommending the right books for you.

Joni: I think you could write a Trump book or a speech and you would not know it was AI.

Shane: Yeah.

Joni: You know, like, the limited vocabulary and the same phrases again and again, I feel like it’s doable.

Shane: It does. It has way too much confidence. It says, “Oh, in 1841, there were 96 wolves in Kentucky that ate 5 people,” and it’s like, “Wow, really?” It could just write…you just believe it because it has a lot of confidence, but it has no concept of truth.

Joni: Can you feed it a plot…like, if you wanted to write a fictional novel, is there any way to give it like, “Okay, I want it to be…” I don’t know, follow this romance trope, for example, or is that too far?

Shane: Yeah, certainly, they call it kind of prompt engineering where, you know, it is trained on the whole internet, but it can be trained on dozens and dozens of romance novels and blogs, hundreds of romance novels and blogs, as well. And so if you start typing the tropes from the romance novel, it’ll complete those. I’ve had an idea of as well because since I make search engines, I would like to make an AI that just writes a ton of stuff pre-written, and then it can be tagged, you know, romance. it has millions of romance writings already done, and you just type romance and find this and that. I think there’s…yeah, so following those tropes, making sure that you hit them. Like, if it’s a teen drama, make sure there’s a trope about the prom in there. I think you can make a system that analyzes your writing and decides whether you’ve hit all the common tropes for it.

Rachel: Do you think that kind of, like, writing tool is kind of what’s next for AI and writing and publishing? Or what direction do you see it going in next?

Shane: Yeah. I use writing tools like ProWritingAid and, you know, a lot of people use Grammarly. And Microsoft just purchased GPT-3, and it’s the advanced writing algorithm. And they’re obviously making a bunch of their own, and so I think eventually, these things will be like built-in to your editors. So, it’ll be in Microsoft Word that can analyze your writing for these things. A lot of it’s, I think, just going to be built-in tools that you already use. The AI is just gonna, you know, since it’s not too hard to include, you just need to create a product out of it. And Sudowrite is one that they are doing a lot of the more advanced prompt engineering and, you know, you choose what kind of style you want to write in.

Joni: Yeah. So, I mean, none of this stuff is replacing human editors or human translation or anything else like that, it’s just sort of a tool to…I don’t know, I guess, a tool to assist you on those things.

Shane: Yeah, I think they would like to. ProWritingAid, you put your whole manuscript in there and it analyzes for you, and certainly, you can correct a lot of stuff before sending it to your editor. But, yeah, developmental editing in that creative aspect, you can’t really replace, you just have to act as the director in a way. So, it can keep writing, but you as the director decide what to include in it. I think that’s the future of it. It may do more of your writing for you, but the creative aspect is still on you to decide where the story’s going to go.

Joni: So, it’s not replacing humans “yet” is where we’re at, we’re good for now.

Shane: Yeah. Well, it’s weird because I don’t think it was ever quite expected to get this good in the first place, and it’s fairly early on. So, I think a lot of human jobs will be replaced by copywriters and just people who write basic blog posts and that. And so I think you do have to lean into what makes you unique and the deep research that makes you interesting, and more accurate, and more thorough than the AI can be.

Joni: Yeah, that’s a good point. I can see a lot of the content farm-type sites absolutely using it. If you’re looking for keyword-heavy and SEO optimization, then, yeah.

Shane: Yeah.

Joni: Interesting. So, what can listeners and readers expect from you next, do you have anything in the pipeline?

Shane: I’m trying to upgrade the system so that it can…because every year, every six months, somebody releases a more intelligent version of it, that they trained at an expense of millions of dollars, and they’ve released it on the public, you know, Google and OpenAI. And so I’m just trying to upgrade the intelligence so that my next book maybe contains more of the writing. This one’s maybe only 10% to 20%, but it would be interesting if it’s actually good to have it do more of the writing. But also, yeah, I’m not sure where I’m trying to go as an author at this point, so I need to decide what to do it. I just read Norm McDonald’s “Based on a True Story.” And it’s such a good book, it kind of inspired me to maybe take a different direction and try to write something more fictional.

So, I’m still working on it. You know, I’ve got the….putting these two books out during…in about a year was probably too much work as well. Having two kids, it was quite intense. And so I’m going to take the next one a little bit slower. I’m doing some stand-up comedy, and maybe we’re gonna have the AI write some jokes for me that I can try on stage. And so just kind of trying to have some fun with it and take it a little slower. That’s my plan for now.

Joni: Awesome. And that segues us nicely into our rapid-fire book questions. Can you tell us what was the last book that you read and enjoyed?

Shane: Yeah, “Based on a True Story” by Norm McDonald. It’s about his life and, you know, being, you know, the guy on Weekend Update on “SNL.” But I think only half of it, maybe less than half of it is true, which is just…I loved because he’s starting to talk memoir, but then he completely makes up the next part and just adds fiction to his life. So, I thought that’s a brilliant book. So, that was it. I read a book by TD Jakes, “Don’t Drop the Mic,” that was a good one as well.

Rachel: And what book is next on your to-read list?

Shane: I picked up some comedy writing books, the creator of “The Onion,” “The Comedy News Source,” he wrote at least three books on how to write comedy. So, I read a lot of books like that, basically, of, you know, sort of, self-improvement and how to do things, it’s the only way I learn. That’s my next plan.

Joni: And did you have a favorite book as a kid?

Shane: Yeah, “The Hitchhiker’s Guide to the Galaxy.”

John: Okay. A big one.

Shane: Yeah. Douglas Adams.

Rachel: All right. Last question. Do you have a favorite fictional AI in books, movies, TV?

Shane: Oh, yeah. Marvin the paranoid android from “The Hitchhiker’s Guide to the Galaxy.” The fact that the robot got so intelligent that he just became depressed and anxious all the time. So, maybe that’s the future of AI is the…you know, like, he was 50,000 times smarter than a human, yet he could barely get himself out of bed in the morning.

Joni: That’s a little sad. But yeah, it probably is. Well, this has been great. Thank you so much for doing this.

Shane: Yeah. Thank you, Joni.

Shane: We will include…so you sent us some images that will definitely put into the blog post, and your website is shaneneeley.com?

Shane: Yeah, shaneneeley.com There’s a lot about the book @stoneagecode.com. And I’m all over the internet as @chimpsarehungry.

Joni: Love it.

Shane: This was a great chat. Really glad you invited me on, and yeah, go KWL podcast. This is cool.

Joni: Awesome. Thank you so much.

Rachel: Thank you so much.

Shane: Thank you.

Rachel: Thank you for listening to the “Kobo Writing Life” podcast. If you’re interested in picking up Shane’s books or learning more about the AI writing he does and machine learning, we will include links to his website and both of his books on the blog. If you’re enjoying this podcast, please be sure to rate, review, and subscribe, it helps us out a lot. And if you’re looking for more tips on growing your self-publishing business, you can find us @kobowritinglife.com. And be sure to follow us on socials, we are @KoboWritingLife on Facebook and Twitter and @kobo.writing.life on Instagram.

Joni: This episode was produced by Joni Di Placido and Rachel Warden. Editing is by Kelly Robotham. Our theme music is provided by Tear Jerker. And huge thanks to Shane Neeley for being our guest. If you’re ready to start your publishing journey, sign up today at kobo.com/writinglife. Until next time, happy writing.

%d bloggers like this: