The lie about machine translation making translators obsolete

In his book Ten Arguments for Deleting Your Social Media Accounts Right Now, Jaron Lanier uses translation as example of how the current free services/ad-based Internet model is not serving us well.

The services we get aren’t the good, he says, and the data we pay for them with is used to spy on and manipulate us. Plus, it threatens to decimate many a profession.

The livelihood of translators, for instance, is threatened by data-driven machine translation (MT). But Lanier points out something usually ignored when people talk about MT. MT has got as good as is it is largely by using our data. I.e., the data we give away in return for and when using free Internet services, plus anything of ours published on the Web (all fodder for the omnipresent Google).

And we’re not just talking about the work of professional translators. Many other types of texts, including email, can apparently be trawled to feed the machine.

What’s more, MT will always need our data. Just to keep pace with human language, which evolves every day, not even to get better.

Lanier’s point is heartening. I’d assumed that truly excellent MT was much further off than the technophiles tell us. Mainly because everyone writes differently. We each have our quirks that must surely put a wobble in any translating machine’s stride.

But what do I know? Certainly not much about MT or intelligent machines.

Lanier, however, is an IT expert. And he implicitly assumes in his argument that MT will always need human input to keep up with the language we humans like to misuse and repurpose.

So MT threatens to make us obsolete with our own data that we keep feeding it. (Just as other data-driven technologies threaten to thin out other professions, like journalism, medicine and nursing, also mentioned by Lanier.)

Our obsolescence is inevitable, we’ve come to believe. But Lanier says it’s not. The problem is with the current Internet model; not the Internet or the technology itself.

What alternative model does he propose? Something like Netflix. You pay money (not your data) for better services that aren’t used by advertisers to manipulate you.

Under this Netflix model, people would be paid for any contribution they make to the Web — articles, code, videos, translations, etc. We have the technology to tag and track contributions. So it is feasible.

Paying users for their work, Lanier says, would also boost the quality of everything on the Web. MT included.

Sounds a bit fair-fetched, doesn’t it? At least compared to where we are today. But who knows? Maybe some of it’s possible.

Lanier’s alternative at least puts the current model of the Internet into perspective. Although obvious, I’d missed the irony that we, the crowd, including but not just professional translators, have fed and keep feeding the big bad MT monster, whether we realise it or not, just by using the Internet in its current form.

Written by Rob Lunn

Rob Lunn is a freelance translator based in Spain. He translates from Spanish and Catalan into English and specialises in legal translation.

Leave a Reply

Your email address will not be published. Required fields are marked *

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