Google has just announced a mind boggling statistic: in a given day, their Google translate service translates in a single day roughly as much text as you would find in one million books. Wow!
The English translation service started in 2001 with the then state-of-the-art commercial machine translation (MT) but the limitations of translation quality forced them to look elsewhere. In 2003, they veered to a new approach to language translation, i.e. learning from data. Basically this approach used Google’s massive computing infrastructure and ability to crunch large web-based data sets to deliver quality (The NCSA at ILLINOIS has done incredible work in this regard as well). Google focused on improving the speed of translation first. In a year, what took them 40 hours and 1,000 machines to translate 1,000 sentences improved to translating one sentence in under a second. Other languages were subsequently added to the translate repertoire. Now 200 million users regularly use this service in 64 different languages. Now that Google has achieved speed, the focus will be on enhancing quality by understanding the semantic content of complex text.
What does this mean for human translators in the future? A casual browsing of the Web as to what people say about the current Google translate program indicates the following complaints: it doesn’t do a good job with translating metaphors, when it comes to translating complex thoughts, Google produces pure garbage, computers cannot understand nuanced language and figure of speech etc., The list goes on and on. The message is unmistakably clear. The translate program has reached its ceiling in terms of effectiveness.
I do not know what the future portends. But I want to caution naysayers that these sorts of reactions have preceded many major artificial intelligence-based (AI) innovations in the past.
In 1997, IBM’s Deep Blue computer did beat the then world chess champion Gary Kasparov (Kasparov alleges cheating on IBM’s side but IBM denies these allegations). A year ago, in a game on T.V. show Jeopardy, IBM computers beat Ken Jennings, famous for winning 74 games in a row. The IBM computers, of course, had access to every bit of information in the world. But the interesting part about Jeopardy was that IBM computers were able to untangle convoluted and opaque semantic statements in shorter periods of time than one of the very best human competitors. In short, these IBM computers were fast thinking machines.
Or think of driverless cars. News reports put that driverless cars have logged an impressive 200,000 miles so far. More impressive is that a legally blind person could drive to a nearby fast-food restaurant by himself. But, as recent as 2004, skeptics warned that driverless cars were impossible because computers would never be able to process a constant stream of information that drivers are exposed, make sense of all these information bits and then make informed decisions. Look at where we are in eight short years.
What does it mean for executives? When it comes to evaluating nascent technologies, we appear to fall into the trap of comparing apples and oranges. We’re usually comparing mature technologies (e.g., driving ourselves in a car) with nascent technologies (e.g., driverless cars) on the very same attributes. Oftentimes, we discount nascent technologies because they cannot do the things that we’re accustomed to doing using mature technologies. What we often underestimate is that these nascent technologies will continue to keep improving and at some point, will be able to match consumer expectations of what these technologies must be able to do. History is replete with examples: online newspapers, online travel websites, in-home pregnancy tests, home blood pressure monitoring etc. (see Christensen’s theory of disruptions for a theoretical exposition of these ideas). At some level, we need to always assume that implausible technologies are soon going to be realities. That mantra should help us in the long-run.
Professor of Business Administration and
James F. Towey Faculty Fellow and
Executive MBA Academic Director