The Rise of the Machines ?

"HAL", from Kubrik's "2001": "Don't worry, Dave, I am just here to help".

A lot has been written in the past few years on the rise of machine translation (computer translation without human intervention) and Google Translate and tools like it. The question is sometimes posed of whether the machines will replace the human translator.

I am not an artificial intelligence expert by any means, but as a translator, I believe that this won't happen, for the following reasons:

• Machines don't think, they match
• Change in AI is only incremental, not revolutionary or a breakthrough
• Translation is about more than matching linguistic codes or signs (please see my blog post about "de-coding"), it is about understanding the real world, and computers can't do that
• Even if the computer is "quite good", you would still need a human to check its work
• One word mistranslated is like a needle in a haystack and can really screw things up

Let's tackle each of these individually. Machine translation actually started in the 1950s, when the CIA had vast amounts of Russian source text but not nearly enough native speakers of Russian to translate it. The CIA then thought that it would be great to just create computers to "process" the information (as if human language translation was a mathematical problem to "solve"). It was a bit naive, and was typical of the "numbers-crunching", behavioralist mentality of the U.S. government at the time (the same mentality that thought the U.S. could win the Vietnam War by counting the number of enemy killed and using various metrics for that). 

The original idea behind AI was to have computers mimic human thought and really "think". As the decades went by though, this became less and less the goal, and now AI and machine translation is just about matching words, using complex algorithms. Thus, machine translation, in my view, does not operate that differently than a CAT (computer-assisted translation) tool, which just matches source text with what has already been translated. I actually don't see a huge difference for instance between a CAT tool and machine translation, other than that the CAT tool is a bit less "wild" in its translation (some people have written that Google Translate translates like it is "drunk"). 

Machines don't think. Even "Watson", the IBM game machine was purely a matching program, like a chess program is. (By the way, the notion that computers beat the best humans I think is a myth perpetuated by IBM. When "Deep Blue", the IBM chess computer, beat chess grandmaster Gary Kasparov about 15 years ago, it was a bit of an anomaly [and Kasparov also claimed indirectly and not so indirectly that he thought that IBM had cheated by helping the computer, by feeding it answers from real humans]). Kasparov immediately asked for a rematch, which IBM declined. Not only did IBM not give Kasparov a rematch, but they tore down "Deep Blue", and have never used it in competition since. Does that sound like a resounding loss for mankind to you ? 

As David Bellos points out in his great book on translation, "Is That a Fish In Your Ear ? Translation and the Meaning of Everything", Google, when it translates its own marketing materials, does not give them to Google Translate, its own machine translation tool, to work on. Google instead hires human translators. Franz Ochs, the head of GT, states that he does not feel that machine translation will replace the human any time soon. Change is only incremental. To take AI from 70% (accuracy) to 90% is vastly easier than going from 90 % to 91 %. 

To understand translation, you have to understand the real world. It is insufficient to understand just language. Language mostly relates directly to the real world that humans live in. 

I just did a legal translation and compared some of the text I translated with what Google Translate did. Overall, I found Google Translate only would be good if you wanted a very rough approximation or "gist translation" to understand what the text is about. But GT would be very dangerous if you tried to use it as a precise translation (for example, a patent filing or court filing, or in front of clients). It confused things with people (consulting vs. consultant), it changed the meaning of things, and missed meaning, and it did not understand the difference between a court judgment and a verdict, and it mistranslated the phrase "not following the law", with contempt for the judge ! (the translation in effect was nonsense and made no sense in the real world). 

Steve Vitek has a good essay on how he feels that machine translation will not only not replace the human translator, but will actually result in more work for us human translators. You can read his essay here: http://translationjournal.net/journal/57mt.htm.

Steve writes: 

"I can only speak about my own experience, of course, but as a human translator who has been translating patents from Japanese, German, French and other languages since 1987, long before cheap or free MT became omnipresent, I believe that I do have relevant experience. I think that it is obvious that MT has not achieved what many people thought it would have achieved by now, namely to put human translators like me out of business. In fact, I think that MT has had precisely the opposite effect: instead of taking work away from experienced human translators, it created more work for them. Because most patents available in complicated languages such as German or Chinese or Japanese can now be translated into English in a few seconds with a few mouse clicks for example with Google Translate, many more patent agents are now aware of what is in these foreign patents than a decade ago, and some of them then in fact do decide to order a human translation based on the information which is provided by Google Translate or another machine translation tool."

and:

"Machine translation will probably never replace human translators. I use machine translation several times a week, usually when I translate a fairly recent patent from Japanese to English, basically because I am too lazy to look up technical terms in a dictionary. When I translate a Japanese patent published after 1994, I automatically print out a short English summary, about two hundred words written by a human translator, and a hard copy of the MT product which I use instead of a dictionary and also to make sure that I did not skip a line or a paragraph, which can easily happen to a human translator who deals with highly repetitive passages."

In other words, at most, machine translation is an interesting tool for us humans. 

Far from a kind of "Matrix-like" takeover, I envision the new world of machines as being good tools that help us translators work better and get more work. (Brian Christian, in his book about AI, "The Most Human Human", states that machines, as soon as you don't touch them, go into "sleep" mode. Far from wanting to "take over", they really just want to rest and conserve energy).