In the book, “The Hitchhikers Guide to the Galaxy”, the author told us about a small, yellow life form that would instantly translate alien languages when inserted into the listener’s ear. While we do not currently possess an alien being with such powers, we do have a BabelFish. While machine translation services may not grant us the ability to comprehend every language in the universe, programs like Google Translate, BabelFish, and other proprietary machine translators give us the ability to translate documents into literally thousands of languages almost instantaneously.

What’s more, the translations are getting closer to human-quality translations and not just the simple transliterations we suffered through for decades. The questions remains, however, will we ever achieve technological singularity in the world of translation? Will a machine ever be able to create a culturally and literarily accurate document in the world’s myriad languages (or even some of them)?

An Introduction to the Concept of Technological Singularity

Technological singularity is the phrase used to describe the hypothetical advancement of artificial intelligence (AI) to the point that it is capable of “recursive self-improvement.” In other words, it can identify its own errors, fix them, and learn from itself.

With the power and speed of such a computer or computer network, the intelligence of the AI would compound so exponentially that the resulting super-intelligence would quickly grow beyond the ability of humans to comprehend or control the process.

The achievement of technological singularity would occur when the AI’s advancement and capacity go beyond what a human mind can even conceive.

The name most often associated with technological singularity is futurist Ray Kurzweil, who took the term from the famed mathematician and science fiction author Vernor Vinge. Kurzweil believes that there will be major advancements in AI by 2020 and full singularity by 2045.

Even then, with an advanced intellect in AI, the questions remain: will these ultra-intelligent computers be able to infer context or detect nuance, comprehend a metaphor or pun, or grasp colloquialisms? Will they be able to apply cultural context to language systems, or will emotion and its counterparts remain a strictly human element?

Why You Need to Consider the Machine Translation Dilemma

Machine translation, or computer aided translation, is no longer just a simple word replacement scheme. Today’s programs are complex systems that use statistics and word use variables from vast language libraries to infer context and estimate proper syntax.

According to Dr. Phil Blunsom, a computer scientist and machine translation researcher at the University of Oxford, “Essentially, we are translating using probabilities to find the best solution. The computer doesn’t understand the languages or know any grammar, but might use statistics to determine that ‘dog the’ is not as likely as ‘the dog.’ What we are doing is a larger-scale version of what was done with the Rosetta Stone.”

With machine translations getting more accurate, will they ever be able to replicate a native-level human translator’s ability? Even Kurzweil has his doubts. In a Huffington Post interview, he noted that by 2029 machine translation should be nearly as good as human-quality translation for most information, however, “Even the best translators can’t fully translate literature. Some things just can’t be expressed in another language.”

Machine translation is becoming more and more accurate, but, for the time being, the question a document author or presenter must ask themselves prior to translation is “how important is translation accuracy?” If the general gist of your text or research is acceptable, you can save time and money by using machine translation services. However, if errors in calculations, inferences, contextual elements and the like are important (as is the case with many high-level scholarly and professional texts, documents, and contracts) then you are not saving time or money as a single mistake or misunderstanding could cost considerable time, money, and effort to fix.

Understanding the Current State of Machine Translation

There will be great advances in machine translation over the next 30 years and the texts will come closer and closer to human-quality translation as we approach technological singularity. Whether we ever get there is not really the issue, the fact remains that we are presently not able to comprehend the advancements in computing and computing technology that will have taken place by 2045. Yet, as Kurzweil pointed out, even with these major advances, there will be elements of translation that computers will be unable to perform.

While machine translation is an effective way to translate most documents, if accuracy is an issue, we are still decades away from replacing the human translator. Machines provide consistent text, but they do not have the ability to comprehend what they are translating and cannot self-correct and edit with precision. When a work requires the use of context, cultural, literal, or otherwise; avoid machine translation or at least back up your translation with a human who has native-level fluency for quality control.

Further, machines do not have field-specific knowledge so they may not be able to grasp the ambiguity of a particular set of information or correctly translate new research and ideas since the translation model is based on a collective of common understanding.

When presenting new data or research, there is presently no substitute for a person with native-level linguistic and cultural fluency and field-specific knowledge. This may be an area where technological singularity will make huge strides, but at present we are not there.

Technological singularity may or may not remain science fiction, but the impact the journey has had, and will continue to have, on translation services cannot be overstated. You may already entrust machine translation with documents that do not require exacting quality and these services will become even more accurate in the years to come.

Given the current state of machine translation accuracy, however, it would not be wise to entrust information that requires error-free text, especially when contextual interpretation or new information is involved, to translation services that rely strictly on this technology.

This post is written by Robert Stitt, a content writer with Ulatus.