Google’s Language Translation AI is Nearing Human-Level Aptitude

unbox-185-google-gnmt-ai-featThe world just became a little smaller and more understandable, thanks to Google’s new AI-assisted language translation system capable of near-human processing accuracy.

Dubbed by the company’s AI team as the Google Neural Machine Translation system, or GNMT, the technique greatly improves results of language conversion with a new methodology. Instead of parsing each word or phrase individually, as in previous systems, the GNMT evaluates the entire phrase or sentence as a whole.


“The advantage of this approach is that it requires fewer engineering design choices than previous Phrase-Based translation systems,” according to Quoc Le and Mike Schuster, Google Brain team researchers.

Initially, the new system produced results at the same efficiency levels as the previous methods. Over time, through refinement by its AI, the GNMT proved to be superior both in speed and accuracy when compared to the older systems. In some instances, Google says that the GNMT has approached almost-human levels of translation accuracy.

There’s still a lot of work to be done on the GNMT though, according to the research team. “GNMT can still make significant errors that a human translator would never make, like dropping words and mistranslating proper names or rare terms and translating sentences in isolation rather than considering the context of the paragraph or page. There is still a lot of work we can do to serve our users better,” Le and Schuster noted.

Don’t be in such a rush, Google. The more you guys improve this technology, the closer humanity gets to creating our future overlord, Skynet, like in the Terminator movies. Just kidding, folks!

Source: The VergeGoogle Research Blog

Related Articles

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Back to top button