Artificial intelligence is making waves again.
Authors are outraged because Meta (Facebook’s parent company) is using their books to train its AI systems. Because those books are copyrighted, Meta cannot use them without paying the original authors a hefty sum.
How did it circumvent that obstacle? The company is using pirated copies, which is technically illegal. Current evidence suggests that Mark Zuckerberg was aware of the actions of his underlings. Authors are mounting a legal defence against Meta.
Whether the courts will uphold the company’s claims that its use of pirated books falls under ‘Fair Use’ laws remains to be seen.
Some have argued that authors are fighting a losing battle because AI keeps gaining new footholds in publishing. In November 2024, the largest publisher in the Netherlands (Veen Bosch & Keuning) made headlines after revealing that it would use Artificial Intelligence to translate a limited number of books.
Translation is tricky. You can’t just translate books on a whim simply because you want to access them in your preferred language. AI can make that scenario a possibility. It can turn a 500-page German novel into its English equivalent within minutes. But every book becomes automatically copyrighted as soon as an author puts the words on a page, and you can’t translate copyrighted content without permission.
Many authors strike book deals that sell the English rights for a manuscript to a publisher. For instance, you can give an Australian publisher the right to sell your English-language novel in Australia.
But if a publisher in Spain wants to release Spanish versions of your novel, they must pay you for those rights. In other words, an author can sell separate publishing rights to every country in the world, multiplying their revenue exponentially.
This assumes that you did not make the mistake of selling all global rights to the first publisher that offered you a book deal. Translating a book for commercial purposes means reaching out to the author, agent, or publisher, whoever has the rights to the language you want.
However, one assumes that readers have the freedom to translate copyrighted books they bought legally if the goal is private consumption. But in such cases, you wouldn’t want to rely on AI systems. Languages are complex.
They have informal expressions, idioms, cultural references, and slang that either disappear or become horribly twisted in a direct one-to-one translation. This happens to me all the time with direct English/Luganda translations. Whenever someone says ‘Well done!’ as a greeting to me, they always get a blank stare.
What does that even mean? ‘Well done’ in what sense? Well done in life? Well done at work? How do they know I have ‘done well’ to begin with? I still don’t know the appropriate response to that phrase.
It took me years to realize that ‘Well done’ is an attempted translation of a Luganda greeting (gyebaleko). A human translator can bypass such confusion by making sure the original spirit of the author’s words comes through.
But that means not doing a perfect one- to-one translation. For now, the best AI can do is clunky one-to-one translations. Therefore, if this concept has ever crossed your mind, I would discourage you from using AI systems to read foreign language books.
katmic200@gmail.com
