Machines on the march

Machines on the march

This content has been archived. It may no longer be relevant

Skill and expert knowledge are crucial for the future

I have an agency client: I don’t want to point any fingers so I’ll just say it’s a client from another EU country, not the one I live in (Spain) and not the one I come from (the UK). This year, as I’ve become a little expensive for many Spanish agencies, I’ve been doing more work for this agency, which has become one of my biggest clients. Anyone who’s read some of my previous posts before will know that this is a big danger sign for me. I always say that when a client becomes my most important customer, it are only a step away from being an ex-customer, but for the moment, I’ve been quite happy working for this particular agency. It pays reasonably well, more or less on time and the project managers are generally pleasant to deal with.

If recent indications are anything to go by, though, this will not last, and it’s not my superstition talking. The client has an e-newsletter for its freelances, and last month’s has set my alarm bells ringing. I don’t want to quote the newsletter directly because it isn’t a public document, but the gist of it, amid the friendly chatter and news of the comings and goings of the agency’s staff, was that end clients are beginning to demand machine translation. For certain types of text and audience, not only is price more important than quality, the end users of the translation are actually beginning to say: “Just run it through the machine and get someone to tweak it a bit”.

And customer demand is what makes machine translation a real threat. While it was just a case of bottom-end agencies trying to con customers and ending up with poor quality results, I wouldn’t have been concerned. It’s far more worrying to find that, although the actual quality may not have changed, clients out there in the real world have heard of machine translation, they know it’s cheap and they believe it’s not too bad.

The upshot of this is that my friendly and by no means bottom-feeding European agency is losing customers and understandably feels obliged to do something about it. So, it’s going to experiment with machine translation, picking a regular client and starting to do all its translations with a machine translation engine and post-editing. Things, of course, are not going to end there. No one says so, of course, but the upshot is inevitable: the translation work I’ve been doing for that agency will be virtually all gone within a year or two.

Nor do I see myself as a post-editor. Some might take to this new “challenge” (as my agency friends would put it), but it isn’t just that the work doesn’t appeal to me, I don’t think I’d be a very good post-editor at all. The problem is that it’s work that seems easy – hence the poor rates it commands – but in fact it’s not at all. Just speaking from my own experience, a good number of the errors that occasionally creep through in my translations are due to what could be described as post-editing failure – when MemoQ translates a segment for me with a mistake in it and I fail to spot it. And then there’s the fact that whenever I’ve been faced with anything like real post-editing I’ve been unable to resist the repeated temptation to scrap the computer’s version and write my own. But if you’re only being paid post-editing rates that’s exactly what you can’t afford to. You have to spot the glaring errors and fix them but ignore issues like style because you’re not being paid to care about that. The more you ignore, though, the more you’re likely to miss something important. No, it’s not for me.

What my customer is telling me is in line, though, with an undercurrent I picked up at the ELIA conference in Barcelona earlier this year. Most agencies at this European language industry conference were cagey about admitting it, but there is a definite feeling not only that the time will come for machine translation but that it will come very soon. In fact it’s almost here. I believe we’re going to have to face the fact that “ordinary” translation work is going to disappear very soon indeed. When I say “ordinary”, I mean the dull jobs, the routine jobs, the translations we’ve all done from time to time that we know no-one is ever going to read. Of course, these jobs can be a large part of our income, depending on who our customers are but they’re simply not going to be there very much longer.

I think it’s time for us to face the fact that if we do a lot of work for agencies – if we do a great deal of this dull work that has been profitable not because it is well paid but because it is relatively easy – we are going to have to change fast. It will either be a case of biting the bullet of post-editing, or moving completely into work where our ability and expert knowledge mean a computer can never compete with the quality we can produce, however many megabytes it has at its command. I’m doing such a job at the moment – the text for a book of photographs with rather poetic captions – for which I’m confident that no machine translation could ever produce anything better than a laughable travesty. Similarly, no computer is going to do a good job on the historical translation I did last week, or the wine tasting notes I translated with week before. This is the future. It’s going to mean doing perhaps less work because jobs will perhaps be harder to find but better work, as quality will be a key factor, and better paid work, among other reasons because those prepared to pay less will simply go to the agencies and their all-powerful machines.

I’m lucky. I’m sure I’m on the way to the right place because there’s a good proportion of my work a machine wouldn’t be able to do. A good proportion, but not enough. Not yet.



  1. Allison Wright

    Thanks for marking this place in time with your blog, Simon.
    I worry about the day when agencies contend that ‘the machine is right’ and the human is wrong, in cases where, in fact, the reverse is true.
    The internet is already populated with countless erroneous (machine-generated)translations. Are we, as human translators, going to be bowled over by big data? I hope I never see the day when MT output becomes the quality benchmark, as opposed to now, when I think human quality still wins the day in terms of accuracy.

    • Simon Berrill

      Thanks for your comments, Allison. I don’t think there’s any doubt that, in terms of quality, humans win the day. What scares me, and what prompted me to write this blog, is that there are now clients out there who think that slightly tweaked machine quality is acceptable if the price is cheap enough. That’s why we need to get a long way from that market quicker than I thought. Only makers of fine furniture will survive the rush to IKEA.

  2. Karin Sander

    Well said, Simon. I also think more interesting work will remain for us, humans. Still, I see the efforts we have to take to get there and to stay there.

    • Simon Berrill

      Thanks for your comment, Karin. I don’t think I’m there yet, but I’m going in the right direction, and that puts me in a better position than a lot of people I know.


Submit Comment

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