Language Technology Isn’t Getting Better, Our Expectations Are Just Getting Lower

Diane McCartney just published a report on Kevin Lossner’s blog about a localization conference she attended (by the way, read the whole thing, since it discusses a lot of very current issues). Among the events, she describes a seminar by Jaap van der Meer, a language technology specialist who has made the (highly revealing) observation that:

MT has not gained the place it has today because computing power has become cheaper and content has been exploding, but because WE the users have changed: we no longer demand fully automatic high-quality translation, but fully automatic useful translation. We accept poor quality because we need real-time translation of even the most trivial piece of information.

“We accept poor quality.” This is the exact point in which the localization industry intersects with the Wal-Mart ideal of instant junk made cheaply. The discourse of seers such as van der Meer is usually Protean, as difficult to pin down as trying to nail Jell-O to a wall. But sometimes they let their guard down and let slip a pearl such as this. I highlight it because it fits like a glove as an illustration for Jaron Lanier’s criticism of contemporary artificial intelligence. The contention is basically that since pure artificial intelligence has failed, we now have to relativize concepts such as human intelligence (or degrade it, to use Lanier’s phrase) in order to whittle down the yawning chasm that still exists between the brain and the output of an algorithm:

People degrade themselves in order to make machines seem smart all the time. Before the crash, bankers believed in supposedly intelligent algorithms that could calculate credit risks before making bad loans. We ask teachers to teach to standardized tests so a student will look good to an algorithm. We have repeatedly demonstrated our species’ bottomless ability to lower our standards to make information technology look good. Every instance of intelligence in a machine is ambiguous. (p. 32)

The problem, you see, is not that technology is improving, but that it isn’t. I am not a Luddite. I am just a disappointed techie. Ten years ago, I truly hoped that massive translation memories would alleviate some of the drudgery of translation and free up resources. Instead, a giant detour was made through MT-Land, in which TMs are simply fodder that is chopped up into tiny pieces for the benefit of the engine. The results are yet to be seen, but so far the marriage between the social media bubble and l10n drivel has been far from productive. Lanier’s unease stems from the fact that—faced with the elusive challenge of replicating human intelligence—the tech industry is forced to redefine the human in order to make it seem more machine-like and dissolve the individual within the crowd. And that has both humanistic and economic consequences. But in the end it is just a rhetorical exercise. The Great Stagnation that economists are talking about is well and truly upon us. Crowdsourcing won’t solve that.

Miguel Llorens is a freelance financial translator based in Madrid who works from Spanish into English. He is specialized in equity research, economics, accounting, and investment strategy. He has worked as a translator for Goldman Sachs, the US Government's Open Source Center, several small-and-medium-sized brokerages, asset management institutions based in Spain, and H.B.O. International. To contact him, visit his website and write to the address listed there. You can also join his LinkedIn network or follow him on Twitter.