No, this isn’t the title of a movie script featuring sentient machines.
Systems Failure
But, in many ways the dawn of this dystopian future, unwittingly ushered in by state actors, is already upon us.
Not all translation will be perfect
Consider a recent article that made headlines around the world: Police use of Google Translate 'mistake'
Danish police arrested an innocent man on some incredibly serious charges based on machine translation (MT) of a text message he’d sent:
But it is even more serious in a case involving allegations of terrorism, and in which the accused are being held on remand
- Thorkild Hoyer, suspect’s attorney
State actors -law enforcement, immigration authorities, etc -employing these kinds of substandard tools in discharging their responsibilities is deeply troubling because of the potential that exists to violate the rights of the individual.
For the uninitiated, Google Translate is statistical machine translation. Essentially, a mathematical algorithm is employed to estimate the probability that a word in one language, say English, will be translated into any particular word in another language, such as Japanese. Contrary to the wild claims of automated machine translation vendors, language professionals such as translators, linguists & academic researchers using a more rigorous evaluation standard peg accuracy at a dismal 30%, give or take a few points. Obviously, this makes Google’s wild claim that “not all translation will be perfect” quite perplexing!
Alarmingly, now the Immigration and Customs Enforcement (ICE) of the United States Department of Homeland Security (DHS) seeks automated phone translation
What is ‘automated phone translation’ and how does work?
Automated phone translation is actually automated phone interpretation. Convoluted use of the term is understandable given that machine translation underpins this technology. However, misuse of terminology is in itself symptomatic of a much larger and more serious issue: that is, a lack of real understanding of the limitations and, therefore, appropriate applications, of these nascent technologies.
Here's an explanation of how phone interpretation works, using English-to-Japanese translation as an example:
- First, speech recognition technology is used to covert speech into text in the source language (i.e. English).
- In the next step, this text is translated with machine translation into the target language (i.e. Japanese).
- And, finally, the translated text is converted back to speech in the target language.
This is a simplified explanation of highly complex technologies. However, the primary technology -machine translation- is, and will continue to be for the foreseeable future, fundamentally flawed.
Why is automated, or machine, translation such a bad idea?
Because at its core machine translation is nothing more than simply a statistical probability of accuracy -- The news article above speaks volumes for the efficacy of this technology.
Experts put machine translation capable of natural human speech at least 30 years -a generation- away! This is primarily due to the fact that:
- We don’t really understand how the brain functions, and therefore are incapable of replicating its functions, including language learning, in artificial environments
- We still don’t understand many of the basics of language yet, such as how children learn languages; therefore, we simply don’t know how to program machines to learn languages
Thus, when a state actor such as ICE relies on these nascent technologies, it doesn’t require a brain surgeon to realize how a compounding of errors will lead to unlawful rendition, as seen in the article above.
How did this dystopian future creep up on us?
Serious research into the possibilities of machine translation started in the 1950’s. And, in the last decade, the field has made some progress. However, this progress has largely been due to massive advances in raw computing power and application of new methodologies such as cloud computing, and NOT due to advances in machine translation itself per se. Misunderstanding of the nature of this progress has in turn spawned a wide misunderstanding of the purpose and limitations of machine translation technologies.
Cost, too, clearly is a factor that continues to fuel the demand for machine translation. In an era of budget deficits and sequestration, quality has unfortunately been superseded by economic necessity.
However, this ill-conceived scheme by ICE has the hallmarks of other half-baked technologies and schemes rolled out the door by state actors before ready for prime time. Case in point:
- The F-35 Lightning II fighter which went into production even before the design kinks were fully worked out (as I write the design is reportedly changing several times a day!).
A combination of the misunderstanding of the fundamental capabilities and purposes of machine translation technologies, together with poorly designed cost cutting measures is creating dangerous policies set to undermine individual rights enshrined in founding documents and statutory law.
What state actors can learn from private hospitals
Here is a hospital which gives due weight to the grave responsibility of patient care. Read Hospital translators at the push of a button
The implication of poor-quality translation in the medical field is a matter of life and death, literally. However, hospitals forego machine translation & machine interpretation for another reason -- Cost. That is, the cost of litigation from medical screw-ups due to machine translation errors.
This is where state actors can take a page from hospital practices. The savings achieved from fielding half-baked technologies and schemes will be far exceeded by the cost of litigation. This is a lawyer's dream -- Machine translation errors will result in lengthy and costly litigation. Further, illegal immigrants and other undesirables that should rightly be deported will instead be able to spend longer in the country while their cases are challenged in the courts.
Why should you care whether these immature technologies become the preferred tools of state actors?
The reason is when irresponsible schemes aren’t questioned and go unchallenged, they tend to become the de facto standard, the norm. And, eventually, regardless of where you live, sooner or later, you, too, will be extricable caught up in this dystopian world where machines call the shots…
About the Author
Ivan Vandermerwe is the CEO of SAECULII YK, the owner of Tokyo based Translation Agency Japan based in Tokyo, Japan. Visit SAECULII for the latest professional case studies, articles and news on Japanese Translation Service
Copyright (C) SAECULII YK. All rights reserved. Reproduction of this article is permitted with inclusion of the "About the Author" reference as is (including text links, japan-translators.saeculii.com), and this copyright information. Articles may not be altered without written permission from SAECULII YK.
0 &DISCUSS response(s)