Translating Artificial Intelligence

By Andrej Mrevlje |

It has been quite some time since I put most of my dictionaries aside. I have a huge collection of them. But these days, most of my printed dictionaries are gathering dust on bookshelves. As I write this post, only “Garner’s Modern American Usage” sits on my desk. It is a pleasant read and can hardly be called a dictionary.

As we spent most of our time at computers, reading, taking notes or writing, we — or I, at least — rarely consult the heavy printed editions of the various dictionaries available. As pleasant as they might be to use, searching through printed dictionaries is time-consuming, and you only receive a short, matter-of-fact answer to your query. We have been using Google search and Google Translate as our main dictionary for years now. You just need to open another tab on your screen — or now even an app on your mobile device — and you dive into an ocean of meanings. I never — or very rarely — use Google Translate for entire sentences or whole paragraphs. I usually use it word-by-word, playing with it across various languages in order to grasp the meaning of the word I want to use when I read or write.

This might change now, thanks to something I read last night. I am a bit scared, but also very much intrigued by what I just learned.

While I still read books almost exclusively in print, what I read on the computer comes to me in the form of various newsletters and grids that I create and change all the time. I no longer have my favorite paper (or other news sources) that I go to in the morning. There is no particular time in the morning or evening for news and other reading on the web. So it was from one of those sources in my collection that I received a piece called “The Great A.I. Awakening –  a long article published in The New York Times Magazine last Sunday. If you can spare an hour of your time, you should read it –it might add something to your life.

The piece opens with Jun Rekimoto, a professor of human-computer interaction at the University of Tokyo, noticing a huge improvement in Google Translate while he was preparing for a lecture online. Author, Gideon Lewis-Kraus writes that:

He was astonished. He had to go to sleep, but Translate refused to relax its grip on his imagination.

Rekimoto wrote up his initial findings in a blog post. First, he compared a few sentences from two published versions of “The Great Gatsby,” Takashi Nozaki’s 1957 translation and Haruki Murakami’s more recent iteration, with what this new Google Translate was able to produce. Murakami’s translation is written “in very polished Japanese,” Rekimoto explained to me later via email, but the prose is distinctively “Murakami-style.” By contrast, Google’s translation — despite some “small unnaturalness” — reads to him as “more transparent.”

The second half of Rekimoto’s post examined the service in the other direction, from Japanese to English. He dashed off his own Japanese interpretation of the opening to Hemingway’s “The Snows of Kilimanjaro,” then ran that passage back through Google into English. He published this version alongside Hemingway’s original and proceeded to invite his readers to guess which was the work of a machine.

  1. 1:Kilimanjaro is a snow-covered mountain 19,710 feet high and is said to be the highest mountain in Africa. Its western summit is called the Masai “Ngaje Ngai,” the House of God. Close to the western summit there is the dried and frozen carcass of a leopard. No one has explained what the leopard was seeking at that altitude.
  1. 2:Kilimanjaro is a mountain of 19,710 feet covered with snow and is said to be the highest mountain in Africa. The summit of the west is called “Ngaje Ngai” in Masai, the house of God. Near the top of the west, there is a dry and frozen dead body of the leopard. No one has ever explained what leopard wanted at that altitude.

Even to a native English speaker, the missing article on the leopard is the only real giveaway that No. 2 was the output of an automaton. Their closeness was a source of wonder to Rekimoto, who was well acquainted with the capabilities of the previous service. Only 24 hours earlier, Google would have translated the same Japanese passage as follows:

Kilimanjaro is 19,710 feet of the mountain covered with snow, and it is said that the highest mountain in Africa. Top of the west, “Ngaje Ngai” in the Maasai language, has been referred to as the house of God. The top close to the west, there is a dry, frozen carcass of a leopard. Whether the leopard had what the demand at that altitude, there is no that nobody explained.

A few days later, Sundar Pichai — Google’s chief executive — reported that the company had completed the initial phase of the transformation and that the future of the company would be “A.I. first.” In other words, Google’s product will soon no longer represent the fruits of traditional computer programming.

Lewis-Kraus obviously spent a great deal of time with the pillars of Google Brain — a department that was founded five years ago and where the machine learning was hatched. It all started internationally, and when the gang got together, “[the] engineers spoke Chinese, Vietnamese, Polish, Russian, Arabic, German and Japanese, though they mostly spoke in their own efficient pidgin and in math,” Lewis-Kraus wrote. Most of the core workers were foreigners, and the story of genius Quoc Le — who, as a child, worked on his father’s paddy fields in Vietnam — is very touching. Quoc Le became one of the fathers of the “cat paper” that later determined and reinforced the new trend in Google’s research in A.I.

Previously, artificial intelligence research was based on programs that tried to mimic high-order cognitive tasks like math or chess, assuming that this would be the path that would eventually lead to something akin to consciousness. But there was another notion buried in the graveyard of history that the Brain team considered: the view that computers can learn from the ground up (from data) rather than from the top down (from rules). “A brain, after all, is just a bunch of widgets, called neurons, that either pass along an electrical charge to their neighbors or don’t,” Lewis-Kraus wrote.

It all started from this point five years ago. It is a fascinating story on how our brains are finally being used as the model to develop machines that can learn. First images, then more complicated matters like language. For this reason, the changes in Google Translate are a great indicator of where A.I. research is heading. Many other internet giants like Facebook, Apple, Microsoft, and Baidu are now rushing in the same direction. Baidu, where Andrew Ng — one of the earlier members of Google Brain — works, already employs 1,300 people to do A.I. research. Google is hiring now, but as this fascinating article tells us, it might be a bit ahead of the others.

This story is not only about the evolution and the incredible acceleration of our A.I. technology and knowledge. Intertwined in all of this is the importance of democracy when it comes to intelligence, be it human or artificial.
Though in order to see any radical improvement for minor languages in Google Translate, we will just have to wait. I checked for Slovenian, my native language, and it is worse than pidgin. I wonder if Google will ever actually resolve all of the puzzles that my native language offers.

☆ Support this work via Venmo

Yonder is a weekly newsletter from Andrej Mrevlje that connects global events in the news, delivered every week. Learn more »