Machine translation tools are cooking up some nonsense. Is this a sign for hope?

According to researchers from Amazon Web Services, reported by Vice (RIP), "a huge amount of the internet is machine-translated garbage".

Great news for translators, no?

I think it does give some hope for the human translators of this world. If the bulk of translation data being used by Large Language Models are themselves misinterpreted data created by machine translation, then surely people are still needed for some time to come, right?

There are some caveats though.

It may be too late for some languages

The study says that this baked in gibberish mostly concerns languages which have a smaller data set online. These are what the researchers call "low-resource languages".

To quote Mehalk Dhaliwal, who worked on the research

"Sentences are more likely to have translations in French than a low resource language, simply by virtue of there being much more data in French than a low resource language."

So, if you are a translator working in languages for which a large data is already present online, you need to compete more with machine translation.

Which leads me to the second point.

It's already changing

I feel that, when it comes to AI and the threat of job loss, many people tend to consider it as a black or white, have or have not situation.

"No worries, AI tools have not taken over my source of income."

"Oh, wait, AI tools have taken away my source of income."

It's not always (going to be) like this. For many industries, the fact that AI tools are inflitrating means people may still have work, but they are

• paid less for doing the same amount
• spend a bulk of their time using the same tools which will make their work redundant

Balancing act

Of course, AI tools also give language professionals the chance to expand their skillset, explore how to adapt and in some cases even enhance their work. How just that balance will be remains to be seen.

Cheers!

Read the full research on VICE. Thanks to them for this.