As a company with AI in the name, we are often asked how we use artificial intelligence and what we do with the data fed into Marlowe, our AI.
It’s a valid and important question, especially as outside technologies potentially threaten the creative processes of authors and other creators.
First, it’s important to note that Marlowe uses Classical Artificial Intelligence, which does not have any generative ability or training. No works that Marlowe reads can be used for the training or creation of content.
Instead, we use analytical AI to compare manuscripts or ebooks with our corpus (or database library) of thousands of published novels.
This analysis can be used for:
(1) helping authors and editors improve a book’s chances for success in the marketplace by providing editorial-level feedback and market placement suggestions
(2) providing readers with book recommendations, based on subject matter or writing style (e.g., if you liked this book, try this one)
How do we accomplish this? Our analytical AI methodology and the training model it’s based on focus on how fiction books compare and relate to each other. This enables Marlowe to determine which elements or storytelling techniques make for a bestseller, or, for example, how a group of science fiction novels are similar or different in composition. Marlowe doesn’t “learn” how to write from reading manuscripts; rather, it learns how to make better comparisons among books.
We value human creativity and creation and security, and protect all data and analyzed works with the latest in encryption protocols and security measures. It is impossible to reverse-engineer Marlowe’s data to reproduce a published novel or draft manuscript.
When authors and publishers use our service to analyze their novels, we don’t keep the manuscript, and Marlowe doesn’t use the manuscript to improve its book comparisons without opt-in style permission. Our corpus of books is continually growing, with cooperation from publishers and authors who are supportive of technology improvements in the areas of editorial tools, book discovery and marketing opportunities. All the texts that Marlowe has analysed have been legally obtained – either from publishers or authors, or by purchasing the print works.
As our corpus increases in size, Marlowe gets smarter and makes more assured editorial recommendations to authors and more precise book recommendations to readers.
We believe that AI is a net positive for authors, readers and the book publishing industry — and that it’s important to use these new technologies responsibly, ethically and legally.
If you’re interested in what is and isn’t permitted under the law, there’s an extensive paper on legal parameters for data mining, created by UC Berkeley, that is available here: https://berkeley.pressbooks.pub/buildinglltdm/chapter/copyright-literacies/