Earlier this month, Betsy Reavley and Fred Freeman, founders of the independent publisher Bloodhound Books, launched a new venture: Globescribe.ai. This AI-powered fiction translation service is “democratising the ability of authors and publishers to get their books out to wider audiences,” says Freeman. “We don’t have a desire for AI translation to take over the world. We are simply providing an option for authors where an option didn’t previously exist,” Reavley adds.
Globescribe isn’t the only company taking a leap into the murky territory of AI translation. Earlier this year, Audible announced that it would be rolling out AI translation services in beta, allowing select publishers to translate their audiobooks for international audiences. Bob Carrigan, CEO of Audible, said at the time: “Audible believes that AI represents a momentous opportunity to expand the availability of audiobooks, with the vision of offering customers every book in every language.” Yet these developments have ruffled more than a few feathers in the industry.
As more companies explore AI literary translation, the rapid progression of the technology, and what that could mean for the future, has divided the books industry. For some, like Reavley and Freeman, the prospect is an opportunity; for others, it is a bleak vision of the future.
Ian Giles, chair of the Society of Authors’ Translators Association, is worried by the pace of change. “Every time you blink there’s a new story like Globescribe or this new Audible service, which worries me because Amazon has so much money and computing power at its disposal,” he tells The Bookseller.
“A lot of our beef is not so much with tech progress. It is with generative AI that has been ‘trained’ using our copyright material. I [even] dislike the term ‘trained’ because it suggests that there’s something sentient on the other side.”
In March, the Society of Authors condemned the “appalling” use of pirated books for AI training after it was revealed that Meta had used millions of pirated books to develop its AI programmes. Authors expressed their outrage by gathering at a protest outside Meta’s London HQ on the 3rd April. Demands for better regulation grew more desperate in the wake of a US court ruling that deemed Meta and Anthropic’s use of books to train AI “fair use”.
Giles also expressed his desire for stronger regulation, given the “unsatisfying” AI-copyright protections in the new Data (Use and Access) Bill: “Would I like the government to be defending writing professionals? Yes, definitely, but apparently Keir Starmer isn’t getting my mental messages,” he says.
Translators’ objections to AI translation are multiple: that it cannot understand context, tone, rhythm or style; that it will lead to an “impoverishment” of language; that there’s no need to translate every English-language novel in existence; and that it simply doesn’t count as translation.
Giles also explained that AI-translated work often would still need to be checked by a translator anyway. “It’s a grey area,” Giles adds. “It’s not our job to tell people how to translate and they’ll all have different ways of approaching it. Do you have to sit there with a Biro and lined paper? No. But if you’re going straight to [generative AI], that’s not translation – that’s checking the computer’s homework, and it shuts down possibilities.”
Frank Wynne, an Irish literary translator from French and Spanish, describes this process as “post-editing” – editing a translation produced by AI. It’s something he, and many of his colleagues, have been asked to do.
“Somebody will send me the gibberish that has been spewed out and I have to take this unwieldy blob of text and turn it into a book.” It would be “easier and faster”, he says, to “start from scratch”.
Continues…