Digital Innovation Digital Publishing
3 mins read

Done in 40 Seconds: AI translates 350-page book on economics, in less than a minute

NASDAQ-listed tech giant Yandex, which operates as Google-Spotify-DoorDash-and-Uber rolled into one in Russia and a host of other countries, has just announced a first-ever critically acclaimed book translated by AI in under a minute (from English into Russian) and published for a mainstream audience.

It took Yandex.Translate, an AI based translation system run by Russia’s most powerful supercomputer, just 40 seconds to translate all 352 pages of “A World Without Work. Technology, Automation, and How We Should Respond,” the latest book by a New York Times praised author Daniel Susskind.

In this book, which was also included in the best business books of 2020 lists by The Times and the Financial Times, Susskind explores the impact of technology, particularly artificial intelligence, on work and society.

Daniel Susskind is an economist at Oxford University and a former advisor to the British government. In the book, he reflects on the future of work in connection with progress in artificial intelligence and the increasing automation of work once performed by humans. At the same time, Susskind notes that humanity has tired of fearing competition from machines and he dispels misconceptions about robots replacing humans.

The book’s translation from English into Russian was of such high quality that only minor inputs by an editor were needed before the book went to print. Yandex’s technology has been endorsed by the industry whose Russian publisher Individuum has made the book available online and will be releasing it in bookstores this month.

For us, this book has become a very serious test of the quality of our translation software, and we are pleased that we passed this test. Thanks to the Yandex.Translate algorithm for large texts, we received an excellent quality text within seconds. A human translator was not needed, just an editor was enough, though usually these people work in pairs.

Anton Dvorkovich, Deputy Head of Yandex’s Machine Translation Department 

Just a couple of months ago, the Guardian published an article written entirely by AI, and now Yandex has proven that translation from one language into another can be achieved incredibly quickly without compromising on quality. This has important ramifications for the future of work and the translation services market, valued at $39.61 billion in 2019 and  projected to reach $46.21 billion by 2027.

To translate the text, Yandex used a Yandex.Translate algorithm typically used for large texts, for example for non-Russian Wikipedia articles in Yandex Search. The book, one of the most talked-about this year, will be available in print in December from Individuum, and its electronic version is already available to Bookmate subscribers. 

To demonstrate the seriousness of the author’s assumptions, new technologies were used to publish the Russian book. The image on the cover was drawn by Yandex’s neural network and the book itself was translated by Yandex.Translate. The Yandex service has long been using a neural network based on the advanced Transformer architecture, which helps achieve high translation quality. The architecture has in many ways served as the basis for YATI, a new Yandex technology that was announced at YaC 2020.  As a result, the 352-page book was translated in just 40 seconds. 

Felix Sandalov, Editor-in-Chief at Individuum Publishers is also convinced that the experiment was successful: 

“We have been looking at machine translators for a long time, as the quality of their work is growing at a dizzying speed. The automation of such a complex and knowledge-intensive activity may worry us, but we also feel optimistic about it. We believe that the algorithm is not the enemy of the human translator, but an assistant. After all, this translation still needed editing. We were saving this until we found a book whose meaning was perfectly aligned with our message, and in the case of “A World Without Work,” this is exactly what happened. 

“The algorithm coped really well with the original text’s syntax, made virtually no mistakes in its vocabulary choices, and did a job that would have taken a human months in under a minute. As is also the case when a human translates texts, an editor worked on it afterwards, too, but this time we left unedited fragments in the text so that the reader could get an idea of how advanced today’s technology is, and perhaps imagine what heights it will reach in the future.”

As the book’s author, Daniel Susskind, also points out, great changes are coming to all areas of our lives, and the book industry will certainly be no exception.

Felix Sandalov, Editor-in-Chief at Individuum Publishers