In the latest episode of the Creating Better Books podcast, ALLi News and Podcast Producer Howard Lovy interviews Alessandra Torre, cofounder and president of Authors A.I. They discuss how the company utilizes nongenerative AI to provide detailed analytical feedback on manuscripts, helping authors enhance their writing without replacing the creative human element.
Thoughts or further questions on this post or any self-publishing issue?
If you're an ALLi member, head over to the SelfPubConnect forum for support from our experienced community of indie authors, advisors, and our own ALLi team. Simply create an account (if you haven't already) to request to join the forum and get going. Non-members looking for more information can search our extensive archive of blog posts and podcast episodes packed with tips and advice at ALLi's Self-Publishing Advice Centre.
Listen to the Podcast: Authors A.I.
On the Creating Better Books podcast, @howard_lovy features @ReadAlessandra of @AuthorsAi, which utilizes nongenerative AI to provide detailed analytical feedback on manuscripts. Share on XDon't Miss an #AskALLi Broadcast
Subscribe to our Ask ALLi podcast on iTunes, Stitcher, Player.FM, Overcast, Pocket Casts, Spotify or via our RSS feed:
Show Notes
Free Marlowe Pro report ($45 value) to all ALLI members. To unlock, use code ALLI24.
About the Host
Howard Lovy has been a journalist for more than 35 years, and now amplifies the voices of independent author-publishers and works with authors as a developmental editor. Find Howard at howardlovy.com, LinkedIn, and X.
Read the Transcripts to the Podcast: Authors A.I.
Howard Lovy: My guest is Alessandra Torre, co-founder and president of Authors A.I., which, as you can tell from the name, helps evaluate your manuscripts using artificial intelligence.
Hello, Alessandra, welcome to the show.
Alessandra Torre: Thank you. I'm excited to be here, excited to talk AI and writing, which is certainly a hotbed topic right now, and probably for the next few years, but that is the way the world works.
Howard Lovy: About three years ago I featured your story on my inspirational indie authors podcast, but just to remind everybody who you are, can you tell me about your background as an author and how you came to found Authors A.I.?
Alessandra Torre: Absolutely. My name is Alessandra Torre, I write romance under Alessandra. I write suspense under A. R. Torre. I've been self-publishing since 2012. I am hybrid. I'm published with Hachette Harlequin, and Thomas and Mercer, which is an Amazon imprint as well.
I used to say that I've always found my greatest success with self-publishing. I hit the New York Times list seven times with self-published books, but I don't know if I can say that anymore because my books with Thomas and Mercer have really taken off. But in 2019, I was approached by a group of best-selling authors. Authors AI was founded by more than 120 best-selling indie authors across all adult fiction genres, and they were interested in using machine learning and natural language processing for editorial feedback, and that was really how Authors A.I. was founded.
I came on as CEO in 2020, but initially just joined as an investor and supporter, and I've been with the team ever since.
Now we're using A.I., it is not generative AI, it is classic AI for analytical purposes and again for editorial, and also now for book discovery and marketing targeting purposes.
Howard Lovy: So, 2019, that's the Stone Age in AI. So much has happened since then.
Alessandra Torre: It was the Stone Age. People would say AI and they didn't really know what AI was. They certainly didn't know AI's capabilities. Back then, it wasn't a foul word that authors would say.
Howard Lovy: It's not totally a foul word, it depends on who you talk to.
Alessandra Torre: It depends on the context, yeah.
Howard Lovy: Exactly. Really quickly, without getting into too much technical detail, what is the difference between classic AI and generative AI?
Alessandra Torre: Sure. At just a high level, AI is typically analytical, like how you would typically have thought in the past about AI. It's like a super calculator, it can process millions and trillions of bits of data instantly and faster, more accurately than a human.
Then you have generative AI, which is really, for lack of a better term, thinking for itself or creating things based on what it's learned from other things, and generative AI is where a lot of controversy is, especially in the creative spaces. So, generative AI uses large language models, and it creates something, it generates something. Sometimes it generates keywords or taglines for your book or book description, and sometimes it generates scenes or plot lines or that sort of thing.
So, Marlowe, which is our AI, does not generate anything. The only thing it generates is output of data, which is not considered generative.
We can talk about retailers in a while and their rules, but there are definitely two very different camps in terms of AI, and a lot of times when you're looking at running your manuscript through something or exposing something you've created to something, it's very important for you to know whether that company is training for generative AI, which is what most authors have an issue with, versus AI being used in a more analytical sense, which is basically just analyzing that and then creating helpful data or analysis from that.
Howard Lovy: So, you're not using authors work to train your AI, but it's more to analyze their writing rather than try to imitate it.
Alessandra Torre: Yeah. To clarify, our AI and every AI has to be trained on something. We have never used customer works to train Marlowe.
It could happen in the future, authors can opt in for whether or not their manuscript is used to train, or would be used in the future to train Marlowe, but it would never be used to train on a generative AI/ if anything, it would train Marlowe in terms of Marlowe just becoming better at understanding what makes up a science fiction book versus a romance novel. That's the sort of training it would do. It would train it to be smarter and to perform the functions that it's performing at a higher accuracy level.
Howard Lovy: So, it's analyzing your work.
Alessandra Torre: Yeah, it's not training how to write books like yours.
Howard Lovy: Walk us through the process a little bit. How does Marlowe help authors improve their manuscripts? Maybe take us through the process from submission to receiving the analysis.
Alessandra Torre: Absolutely. So, I think it's important to say where Marlowe's roots are, and if you are a data geek or just a word nerd, like many of us authors are, a really fascinating book is called The Bestseller Code. It was published 4 or 5 years ago by Dr Matthew Jockers and Jody Archer.
Dr Matthew Jockers is one of our co-founders at Authors AI, and that was the early precursor of what later became Marlowe. So, Marlowe was initially trained on distinguishing the difference between a best-selling book, by best-selling, it was a book that had been in the New York Times list for more than 10 weeks, a consistent bestseller, versus a strong seller, versus an average seller, and trying to understand if you looked at a thousand books that were superstars, what made them different from a well performing book, and really analyzing kind of the DNA of those bestsellers. It's a really interesting read.
It talks about how you can tell the difference between an author from the UK versus an author from the United States versus how many times they use the word ‘the' in their writing. Just a lot of really interesting little factoids.
I found it really interesting, but that was what Marlowe was initially trained on and what Marlowe still does to this day. If you upload your manuscript at authors.ai you can get a 34-page, overview or analysis of your book. So, you upload your manuscript, you tell Marlowe a few details about your manuscript, the main characters names, whether it's 1st person, 3rd person, that sort of thing, and then Marlowe analyzes it in a few minutes and produces an output.
You can see sample reports on our site. We have a free basic report, which anyone can do as often as you want, and they can tell you things like all of the cliches in your book, a lot of just interesting stats and analysis, but the pro report is where all the juicy and fun stuff is, and ALLi members do get 1 free Marlowe pro report, which is normally $45.
If you're listening to this, if you're an ALLi member, ALLi24 is the coupon code for that, for a single use pro report.
The pro report has a lot of things. It's like looking at your book in a totally different way. You can see where the dialogue exists in your book. A typical 25% to 35% of a bestselling novel is dialogue, so you can see your percentage of dialogue and where it exists in your book, what the plot line of your book looks like, and what bestseller it's closest to, just a lot of really interesting things. Again, it's a very developmental edit, that sort of top level, it is not line by line. It's not a copy editor or anything like that.
Howard Lovy: I'm almost afraid to feed my work into it and see how many cliches I use. I've been told I use the word ‘just' too much too.
Alessandra Torre: What do I use? I use ‘already', there's a couple of words, because it also identifies how many times you use, per 100,000 words, adverbs and adjectives, and there are a lot. ‘Extremely', I use ‘extremely' way too often, and it'll tell you how often bestsellers use it versus you, and it's very interesting.
Howard Lovy: When I'm not podcasting, I'm also a developmental editor. Should I be afraid of this, or should I use it in my work, or both?
Alessandra Torre: The good news for editors is that Marlowe doesn't replace human editors. So, Marlowe can tell you that your book, for example, is slow in the middle, there's a big gap in the plot beats in the middle, but it can't give you ideas of how to fix that. Or it can tell you that your main character is too, I don't know, that all of your characters personalities are too similar, but it's not going to tell you how to fix that.
So, we have a lot of actual editors who use a Marlowe report to support the feedback that they give their authors, and also just to supplement. It's an interesting thing to include with a developmental letter, is a Marlowe report.
But for authors who can't afford an editor at all, I think it's much better than not having anything.
Howard Lovy: We talked about before, a lot of writers are naturally skeptical about anything AI. Are you facing more of those challenges of skepticism from writers who are wary of using any kind of this technology?
Alessandra Torre: It's been interesting. When ChatGPT came out, that really threw a bomb into everything in our world, because suddenly everyone was aware of AI, and for the most part, alarmed by AI.
So, we encounter a lot of skepticism. We encounter a lot of emotions, period, and it's very important that people do understand the difference between generative AI and AI assistance, which is how the retailers define the 2.
For example, when you upload your book on Amazon, it asks if you used any AI software or anything like that, and the terms of use very clearly distinguish the difference between AI assisted programs, like Marlowe, and AI generated content. They don't care or need to know about AI assistance, they do want to know about anything that was AI generated, even if it was heavily edited by you after the fact.
So, we do hear that a lot from authors, that's a question, and then we have a lot of authors that just don't, I guess, don't think that Marlowe works, but that's the beauty of running the free report.
A lot of authors are also a little afraid, I think, of feedback, but I often get emails about Marlowe's very kind and nonjudgmental. It's just black and white. This is the analysis, and it's freeing once you get it and you realize, oh, okay, if I ignore this, no one's going to be mad at me. It's nice sometimes not having the human element involved in the feedback.
Howard Lovy: When people talk about either legal regulation or industry standards, there's this difference between classical and generative AI. If you use ChatGPT, you have to disclose it. If you're using Marlowe, not so much, it's akin to say using a spell checker.
Alessandra Torre: Yeah, you absolutely do not have to disclose it, and there's a block of texts that I refer to a lot in emails, but you can find it just by searching KDP AI assistance. The gist of it is, you can even get plot ideas from AI, not that Marlowe provides those, and that's okay, but anytime you've received any sort of output that's AI generated and used it, including art or graphics, then that's AI assistance.
And I hate to bust anyone's bubble, but Grammarly uses AI. So, there's a lot of times that authors are using AI, and they don't realize they're using it, like with keyword tools and that sort of thing, a lot of that falls under AI and machine learning, and NLP, which is natural language processing.
Howard Lovy: Okay, so that makes me feel a little better using that. Our audience is primarily indie authors, self-published authors. How can self-published authors benefit from using tools like this in particular?
Alessandra Torre: I think the beauty of self-publishing is that we have full control over everything, which is a blessing and a curse because it also means we have to do everything.
As a self-published author, I run a Marlowe report frequently. As soon as my first draft is finished, I run one then, and then I do a couple rounds of edits, and then I run one again, do a couple rounds of edits. For me, it's extremely helpful.
There are authors that absolutely love Marlowe, and there are authors that don't take anything from Marlowe. So, it really just depends on your own personal editing type.
But I think in terms of getting a second opinion and a different opinion that does differ from humans, just in terms of non-biased, and also, I don't want to say clinical, but raw data and the look of the DNA of your manuscript, I think it's a fantastic editing tool.
Howard Lovy: Are there any success stories you can share about authors who significantly improved their work?
Alessandra Torre: Yeah, we have a lot of authors, we have a lot of great testimonials on our site. So, it's important to say that Marlowe is for fiction authors. It is trained on fiction. We do have non-fiction authors; I just had a non-fiction author the other day who sent me a really nice email. So, we do have non-fiction authors that use Marlowe, but really all of its graphs and charts and training is on characters and on plot lines and on storytelling and writing style and that sort of thing.
What we have coming in the future and where we really hope to help indie authors is with book discovery.
Right now, Marlowe's skills are being used on the editorial side, but the truth of the matter is that Marlowe is and can be a super librarian. When Marlowe reads a book, it knows everything there is to know about that book. It remembers every scene, every line of dialogue, every emotion and where it exists in that book, and when Marlowe then reads a million or 2 million books, then if a reader says, Oh my gosh, I really love Lord of the Rings, but I wish that had spicy sex in it and, Marlowe could say, oh, here's a book that's 92% similar to Lord of the Rings in terms of writing style and subjects and pacing, but it has high sexual content.
Or the opposite, if a reader said, I really love the writing style of Stephen King, and I like books set in the Scottish Highlands in the 1800s, what do you have that's in the genre of fantasy? Marlowe can produce that instantly.
Where I would really love is for us to be in a world where indie books are on the same equal, footing as traditionally published books, and stories are being recommended and discovered based on plotline and not based on marketing dollars or advertising dollars.
That can happen with technology, and that really is how we should be recommending books, and how a reader should be finding books is based on the storyline and the writing style, not based on buyer behavior, which can be really heavily influenced by dollar spend.
Howard Lovy: Is that something in the future, the reader focused feature?
Alessandra Torre: Marlowe already has the capabilities for it. Marlowe can already do everything I described. What Marlowe needs in order for us to launch that is Marlowe needs to read as many books as possible. So, right now, that's our next roadmap. Can we create a portal where indie authors can submit their books to Marlowe's recommendation engine, and can we get enough? Because we really would like over a hundred thousand in order for Marlowe to do accurate recommendations. It can only recommend books that it has read.
So, that's in the works now. We've just launched a crowdfunding campaign on Start Engine, and that'll help give us the funds we need to have the staff to support that and to launch that.
I'm really excited, and I hope it comes out before the end of 2024, but gosh, this year has passed so fast. So, I hope it's not three more years before we talk again, Howard, but if it is, I am certain it will be launched by then.
Howard Lovy: So that engine is agnostic when it comes to whether it's traditionally published or self-published?
Alessandra Torre: Yeah, it would be completely agnostic. It wouldn't be influenced by anything except reader filters. So, if a reader says, with all of my book recommendations, I don't want books with any foul language, or if they say, I only want books written in first person point of view to be recommended to me, or something like that. Then a reader could also say, here are my top 20 favorite books, and Marlowe could analyze those, find the commonalities between them, and say, okay, here's 10 more books that are similar to the characteristics that you seem to enjoy. You know what I mean?
And the beauty is Marlowe's common characteristics aren't things like metadata. It's just this reader prefers books that are happy in the beginning and then have some sort of a tragic event in the end, or the reader likes fast paced books with high sexual content or whatever it is, it collects hundreds and hundreds of data points.
So, it's really interesting sometimes to see what similarities exist between books and how books that don't seem that similar are similar when you get down to the DNA of that book.
Howard Lovy: So, this is very specific. It's the Amazon also recommends on steroids.
Alessandra Torre: Exactly. Yeah, if Amazon also recommends was 100 percent transparent in terms of the reader being able to dictate it. I'm assuming Amazon also recommends is influenced by buyer behavior. It's hard to say because it comes and goes, but yeah, that's a great example; Amazon also recommends on steroids.
Howard Lovy: I think your website also mentions Marlowe is being used by academic institutions. So, how are educators using it?
Alessandra Torre: We have a lot of educators, anywhere from high school and middle grade teachers to college classes, and typically they use it just like, they'll be studying a classic book, and they'll use the analysis is an example in their teaching. Or sometimes we have creative writing classes that we give free Marlowe reports to, so they can run analysis on their books and get feedback. That sort of thing. We love academia, the majority of our team are academics. So, our roots are in, of course, books, libraries, and academia.
Howard Lovy: I'm not sure, maybe this is a loaded question, but what does this mean in terms of originality? It's taking what works and what sells, but oftentimes innovation comes with something so completely different there's nothing else like it out there.
Alessandra Torre: One common concern we used to hear a lot was like, oh, if there's a best seller standard that Marlowe is helping provide feedback for, how are all books not going to be similar to each other? Is that what you're referring to or are you referring to us offering something that's different than what's currently in the marketplace?
Howard Lovy: No, I'm talking about how it analyzes books that are completely different from anything you've read before. Is that automatically a negative because it's not similar to Stephen King, or whoever?
Alessandra Torre: No, I love that question. What's interesting is, if we go back to the bestseller roots. The Circle got a near perfect score on the bestseller, even though it was extremely different than the others.
Because a lot of times Marlowe's analysis, I'm trying to think of an example of 2 books, because a lot of times Marlowe will recommend a book that we're like, that book isn't at all similar to this other book, why would Marlowe think that the same reader would like both of these books, but then it does. Or we'll feed in a book that was a surprise hit. I Robot was an example. I Robot, we thought Marlowe would struggle with because it's written in a very unique way, and nonstandard in its formatting.
So, a lot of times we test it. We test Marlowe's bestseller-ometer. We'll give it a book it doesn't know anything about. It doesn't know the author's name. All it knows is the manuscript, and then we ask Marlowe to say its likelihood of being a bestseller, and Marlowe has a 93% accuracy rate, 93% to 95%, depending on the data set.
We give it some funky bestsellers, some bestsellers that you wouldn't think, and very consistently, Marlowe gets it right. Marlowe gets it right when we oftentimes get it wrong. So, I don't have a way to explain it, really, because we don't always understand it, but it's a certain analytical makeup of different things, and it includes things like, does the plot line follow one of these seven standard kinds of archetypes of novels, which is something that was talked about for decades? It has to have the general things, varied pacing, varied sentence structure, but typically it comes down to word choice, and that is how writing styles are typically identified is the word choice that they use and the frequency and combination of words that they use.
And that Marlowe understands at a much more granular level, times a million, versus me, or a standard individual.
Howard Lovy: Yeah, I've noticed that when people talk about AI, there's a magical thing happening, and I think they call it the black box, where you don't really know what's happening, but it works.
Alessandra Torre: Yeah, and I think it's important to say, especially with LLMs. I talked a little bit about LLMs with generative AI, and Marlowe does not use LLM technology, but one of the benefits of that is LLM is very much a black box. It can tell you these two books are similar, but it cannot tell you why.
It's not that it keeps it a secret, it's just it can't tell you why, it can just tell you that there are thousands of data points that point to that. But what is special about Marlowe is that Marlowe can tell you, like the word choice, it's not going to be able to explain to me without it taking hours of thinking trying to explain to me how word choice is similar between these two books, but what it can say is when I'm like, why is Lord of the rings and Da Vinci code similar? It can say, okay, because in word choice, they're a 92% match, in pacing they're an X percent match. And it can say, and they both have common themes of abandonment, or something like that.
It can break down kind of the recipe of why it said these two books are similar. So, that's very helpful to us, but what it also then allows is for a reader to be able to say, okay, now that I know what commonalities you found across my top 100 books, here are the ones that I really care about, and these are the ones that I want you to recommend books for me in the future. So, that can be enormously helpful and powerful.
Howard Lovy: Oh, that's fascinating. What market trends do you see in the publishing industry regarding the adoption of AI? Are more authors and publishers becoming open to using these tools?
Alessandra Torre: I think more authors and publishers are becoming more educated about the tools. We have been in conversations with every major publisher for years now, and when ChatGPT came out, there was a really big withdrawal from everything related to AI, and now they're coming back with teams and committees that are educated on AI and what training on AI means, and how exactly their data will be used. It isn't like before where we would have conversations and there was a learning curve, or they didn't know the questions to ask. Now, both authors and publishers are coming to the table, much more informed, and that's great, because there is that understanding of the distinction.
Over a year ago we came out with our stance on the use of AI in fiction, and it's on our website if anyone wants to read it, it's in the footer, where we drew a line in the sand and said, we don't train on generative AI, we don't develop anything with generative AI, and that helped us a lot with publishers and also with authors.
As far as trends, I think there's going to be a lot of legislation coming out in the future because we did come into this brand new universe that really didn't have a lot of rules and so there wasn't a lot of rules about how retailers could use our information and our manuscripts, how companies that we uploaded our manuscripts for things like formatting services and things like that, we didn't really have an awareness of how our books were being used.
So, I think there's going to be a push for much more transparency and legal stipulations on the transparency of how works are being used, and also how transparency in the corpuses that are being used to train technologies.
All of Marlowe's training corpuses were either purchased or attained through permission from the author or publisher. We've had a few crises in the last year or two from companies that popped up that were trained on illegally obtained materials. So, I think there's just going to be a lot more accountability, which is a great thing, and a lot more education, and I think that authors will become more comfortable with using AI, and I think we're going to have a little bit of a divide in the culture in terms of, because there are authors that are embracing generative AI in their creative processes, and it is very much a divided room right now, in terms of thought processes on that.
And there's a lot of use of generative AI in marketing purposes that I personally don't think has any ill effect on humans. If you aren't currently hiring someone to write your marketing taglines or your marketing ad copy, you're not really taking a job away from a human if you use an AI to help you with that. But that's a divided issue that another author could have a big issue with, so it's just everyone's personal preference and where they sit on different issues.
Howard Lovy: People are divided in the self-publishing community. Many purists are out there, but I think we have to take it on a case-by-case basis.
Alessandra Torre: Absolutely, and our community is divided on so many things, so it would be strange if there weren’t conversations among people.
Howard Lovy: I wish I could talk about this for longer, but we're just about out of time. My last question, is there anything else that you want to add about AI or your company that I didn't ask?
Alessandra Torre: I don't think so. We'd love you to give Marlowe a try. If you're listening, again, you can use ALLi24, all caps, if you want to try a free Marlowe pro report, or you can try a basic report without using a coupon code. You can run as many basic reports as you like, but we're happy to answer any questions and have any conversations about AI. I love chatting AI. It's been fantastic talking with you today, Howard. So, thank you so much for your time.
Howard Lovy: Thank you so much, Alexandra, and thank you for your insights into AI.
Alessandra Torre: Absolutely. Bye.
