ChatGPT already used to author hundreds of self-published ebooks, as questions remain over ethical, legal issues – not to mention quality
The popularity of OpenAI’s ChatGPT has extended to the book industry, with hundreds of self-published titles on the market generated at least in part by the chatbot.
Amazon’s US and UK Kindle stores listing more than 200 titles with ChatGPT as author or co-author, an indicator of how the tool is being used to create books of all kinds.
The titles range from technical books about ChatGPT itself to children’s stories to a book on anatomy or a compendium of strategies for “Reigniting the Passion in Your Marriage With Immediate Effect”.
Some of the books include illustrations generated by OpenAI’s Dall-E automatic image generator.
The chatbot generates text from humans’ prompts, based on material scanned from the internet.
This has raised questions for the ethics of using it for creating academic papers or published works, since the material generated may be plagiarised from other sources.
It may also simply be inaccurate, as Google demonstrated when an advert for its similar Bard tool earlier this month showed the chatbot confidently asserting false information.
Reuters said it had found hundreds of tutorials for creating a complete book in just a few hours on YouTube, TikTok and Reddit.
The news service found books generated by ChatGPT on subjects including get-rich-quick schemes, dieting advice, software coding tips and recipes.
Amazon, whose popular Kindle Direct Publishing allows self-publishers to sell their own works on the Kindle Store, has no requirement to disclose if an AI has generated or helped to generate a text.
Mary Rasenberger, executive director of writers’ advocacy group the Authors’ Guild, told Reuters there was a need for transparency from such platforms to avoid ending up “with a lot of low-quality books”.
Amazon said in a statement that all books in the Kindle store must adhere to the company’s content guidelines, “including by complying with intellectual property rights and all other applicable laws”.
Aside from legal issues, early experiments with ChatGPT as an authoring tool have shown lacklustre results, according to some users.
Stephen Carter, a Yale law professor and author, said in a recent Bloomberg column that he had found ChatGPT “not bad” at producing plot-driven stories, although its “endings are uniformly weak”.
Its description of a “beautiful sunset” sounded like a “middle-schooler who’s trying too hard” and Carter found that the chatbot often resorted to clichés.
Mark Dawson, who has reportedly sold millions of copies of self-authored books through Kindle Direct Publishing, told Reuters ChatGPT-generated books were “dull” and were likely to quickly “sink to the bottom” as a result of negative reviews.