Can we tell humans from machines? | by Xenda Amici | Unbabel Community | May 2022

I spoke to Unbabel’s Vice President of Language Technologies about “imitation play” and the human behind the machine

If you’ve been online today, chances are you’ll be asked to prove you’re not a robot. You may have thought “oh, not that again!” trying a second time to tag all images containing a traffic light. The next time you see one of these CAPTCHA posts, you’ll have a name to “blame”: Alan Turing. CAPTCHA stands for Completely Automated Public Turing test to tell Computers and Humans Apart, developed with principles created by English mathematician, computer scientist, logician, cryptanalyst and philosopher Alan Turing.

The purpose of the Turing test, described in the 1950s article Computing Machinery and Intelligence, was to answer the question “Can machines think?” by analyzing whether a human interrogator could tell, through conversation, the difference between a machine and a human being. The Turing test became more widely known after the film The Imitation Game (2014), showing the creation of the decoding machine that helped the UK decipher German messages during World War II.

“Instead of trying to produce a program that simulates an adult’s mind, why not try to produce one that simulates a child’s? If this was then subjected to an appropriate course of education, one would obtain the adult brain.
I am Turing

Would machine translation engines pass the Turing test?

I posed this question to Alon Lavie, Vice President of Language Technologies at Unbabel, Consulting Professor at Carnegie Mellon University (Pittsburgh – Pennsylvania, USA) and 2021 winner of the Makoto Nagao IAMT Honor Award for his contributions in the field of machine translation. “In conversational contexts, today’s chatbots come close to passing the Turing test in terms of language generation capabilities… [these days] it largely depends on the language, type of content and context. In some situations, it becomes increasingly difficult to tell. But in most business use cases, we still need humans to ensure consistent, high-quality, on-brand translations. »

Ex-Machina

Since the introduction of machine translation into the workflow of translation processes, one of the biggest concerns of the translator/reviser community has been that machines will eventually replace human expertise. Alon shared with me the two perspectives he witnessed following this fear of translators:

When translators are asked to post-edit machine-translated text, they often find that the target text is so bad that it would take more time and effort to correct it than to translate the text from scratch. Alon tells me that, while that was true a few years ago with earlier machine translation technology, with today’s neural machine translation, the time an editor spends on a machine translation job is is significantly improved for most major languages.

“The new neural models are very good most of the time, and they generally generate fluent and grammatically correct language. But they still make critical mistakes. The good news is that AI technology for detecting translation errors is also rapidly evolving. It will therefore become easier for translators and editors to detect these errors, but we will continue to need humans to correct these errors”.

Nowadays, machine-driven efficiency is growing in many fields, not just linguistics and translation, and many professionals in all fields are wondering how their roles will change because of it and how to better keep up with the rapid changes. in their markets. If we think back to the beginning of the Industrial Revolution, this is not a new concern. Almost two hundred years later, we carry out work in all fields using tools and machines without which we cannot imagine working, even if at the time of their invention we saw them as a threat.

All these tools are meant to do is help us by doing work that can be done by a machine, so that humans can step in where their contribution is irreplaceable.

Alon had this to say:

“I sincerely believe that the translation community will never be unemployed, but their work will be different from what it was in the past. For the most part, translators will become the expert editors who guarantee the accuracy of the target text, the correct semantics of the subject and the brand language of the companies.

Lots of resources 🎁

The Association for Machine Translation in the Americas (AMTA) for academic papers, presentations, lectures, and tutorials has a lot on the latest AI translations; Alon recommends it to anyone who wants to learn more about the subject.

Another great resource is the European Association for Machine Translation (EAMT). We’ll take a moment to mention that seven papers have just been accepted for their annual conference in Belgium! This is our biggest entry yet, so well done, Unbabelers!

Let’s continue this conversation
Do you think the machine translation engines we have today pass the Turing test? Share your opinion with us!

James G. Williams