In the Power of Money-Making Algorithms We Trust




Greater than 5 minutes, my friend!

I don’t really believe very much in the power of algorithms. I just don’t trust them too much.

Unlikely bedfellows though they are, driverless cars, drone killings, and machine translation have one thing in common: they are controlled by algorithms, and algorithms can be lethal. Sometimes, an algorithm can kill by driving a driverless car into a tractor trailer, and while a drone (“believed to have been operated by the US”, as the New York Times put it) is relying on smart algorithms, it sometimes assassinates 60 civilians in a funeral procession.

Machine translation has been so far diligently killing, quite gruesomely, not just our language, but also jobs for translators, although it is quite possible that people relying on information “translated” by an algorithm were already or will be killed one day too.

According to Slator.com, a translation platform called Fluently is going out of business because the company ran out of cash and failed to secure further financing. Fluently was an online translation platform, (one of many) that sought to automate away the project management part of translation. Although the company says that it signed up 2,000 translators, the company eventually did not make it. I think that what in fact killed this (yet another) innovative platform was the company’s excessive faith in the power of algorithms. Here is how Slator described what happened:

“Minimal Human Interaction

In part, Fluently relied on clients’ willingness to manage their own projects, and on translators open to completing jobs with minimal human interaction.

The way Fluently works is clients log in, drag and drop files, and then get an instant word count, translation memory leverage, and a quote. They then go through a briefing screen, where, using a slider, they are able to select, for example, what style they prefer (formal, informal, etc.) and pick the translator’s area of expertise.

After clients choose from among the matched translators served up by Fluently’s algorithm, they are required to pay upfront. Payment is done through the platform and the funds are held in escrow until the translation is delivered.

There was a lot of rejection when a customer picked a translator, but the translator did not take the work. Fulfillment became difficult.”

—Karin Nielsen, founder of Fluently

Most translators will understand instantly that it is not terribly difficult to, “sign up 2,000 translators”. There are so many of them out there! Even somebody like myself, a one-man operation which does not provide a “translation platform”, and which is not really a translation agency either, receives dozens of resumes from translators hungry for work just about every day, and I usually delete without reading them.

Matched Translators Are “Served Up” by and Algorithm

I love the description, “After clients choose from among the matched translators served up by Fluently’s algorithm”. Smart translators are served up by smart algorithms like a meal in a restaurant, huh? But if the algorithms were so smart, how is it possible that ” fulfillment became difficult” because the translators in the end did not want the job?

Because I don’t believe in the power of algorithms when it comes to finding good translators, I only work with a few translators, almost always translators who were recommended to me by other translators; currently less than a dozen.

But unlike in the case of the Fluently platform, soon to be or already deceased, my translators almost never reject a job offer and on the rare occasion when they do, usually because they are on vacation or because they can’t fit the job in (although my deadlines are generally very generous), they usually suggest colleagues who are equally skilled in the art of patent translation and who, bless their hearts, charge about the same.

My guess is that Fluently’s signed-up translators rejected the jobs offered to them for two reasons: 1) because they were poorly paid, and 2) because the deadlines were short.

I should mention that the rates I pay translators who work for me are not really stellar (as one of them snidely commented to me). But still, they are much higher than the rates translators can expect, regardless of their education, credentials and experience, on “translation platforms” such as Proz or Translators Café. (And the rates on these platforms are probably still much higher than what Fluently was offering.) So that I can sleep at night, I pay translators who work for me the lowest rate that I myself am willing to accept for my work from other translation agencies. Even so, there is still a nice profit margin in it for me, too.

I know from resumes that I receive daily that I could hire translators for the translations that I am unable to do myself, (usually due to the wrong language direction), but that I can proofread quite competently since I almost always know both languages. Also, given that I have been in the business of patent translation, both as a translator and as a mini-agency for almost three decades, I know a little bit about how things work in patent translation and don’t need to rely on a combination of databases with very smart algorithms; I can just use my human brain.

I know that I could easily find translators who charge much less than what I am paying when I look at the rates that translators themselves advertise on what are called “language platforms” and that I could contact translators who are much cheaper.

But greedy as I am, I fear that if I go for maximum profit for me, me, me by looking around for translators offering lower rates, I will eventually pick a subprime translator without realizing it, or even though I may pick a relatively good translator, this translator will be under such pressure to accept as much work as possible to pay the bills, that in the end, he or she will start making too many mistakes.

And if I am not able to catch these mistakes, I will eventually lose a client that might have been with me for many years.

Staying Away from Algorithm to Play it Safe

So I try to play it safe. I don’t use logging, dragging, dropping, or “translation memory leverage” which I assume means wage theft through algorithms referred to as “fuzzy matches” and “full matches”, a very popular racket nowadays in “the translation industry”.

Nor do I ask my customers to pay in advance to make sure that I will get paid for the translation even if the quality is so horrible that the customer might balk at paying for it. Well, I do ask for advance payment if I am unable to ascertain that the customer is a bona fide patent law firm or a real company that has an address, a product to sell, and employees.

If I deal with an individual, I usually ask for a down payment of 50%, but this is the only time when I do so.

I believe in replacing algorithms by human interaction as much as possible. I understand that replacing human interaction with smart algorithms as much as possible would certainly be a very profitable concept for a translation platform. The owners of the platform could save a lot of money that would not need to be paid to project managers when everything is done by just dragging and dropping files into automated menus.

All you need now is a few thousand translators hungry and willing to work for what an algorithm determines their work is worth – and you have created the most innovative, ultimate cash register platform in “the translation industry”. Ka-Ching with minimum human interaction.

But minimizing human interaction, profitable as it might be, is not a whole lot of fun for clients, or for translators. It would be kind of like trying to limit human interaction to a minimum during sex. I suspect that fulfillment might become difficult. Human interaction, when it works well, is not only necessary, but also fulfilling, and fun.

It is not an accident that the most severe type of punishment in prisons is a model that is based on minimum human interaction called solitary confinement.

Steve Vitek

About Steve Vitek

Translation of patents from Japanese, German, French, Russian, Czech Slovak and Polish since 1987. Blogs at www.patenttranslators.com, website at www.patenttranslators.com

7 thoughts on “In the Power of Money-Making Algorithms We Trust

  1. Great post, Steve! And I agree with every single word. We’re in the business of human interaction, automating and minimizing human interaction won’t play nice, as a matter of fact, it is often a short track to disaster.

    From the article I could understand that there was an assumption that translators will always be available and willing to work. But from my short experience of finding and hiring translators I’d say it’s not always the case. Good translators are always busy. At good rates too. No wonder nobody wants to take jobs generated by algorithm when you have human clients, who pay you well, who answer your questions, who talk to you.

    I’d choose humans over algorithms any time of the day. And I’m talking both as a client and as a translator. When you’re running an agency-like platform project managers cannot be replaced. Same goes to customer service and finding talent. Those services can only be provided by humans. That’s why platforms that try to automate this will keep failing miserably until they realize that we’re in the business of human interaction and not “serving translators via algorithms”.

    Report comment
  2. Having said that, I believe there’s a room for automation in project management. But for it to work, platform designers have to understand the needs of both their customers AND their workforce (i.e. translators). So far I haven’t seen that many platforms that have the grasp of what translation is all about.

    Report comment
  3. There is always room for automation of probably anything.

    But I believe that when “the translation industry” over-automates the system, it in fact inadvertently gives an “unfair advantage” to sole practitioners who don’t use any automation and instead communicate with clients and with translators directly. Don’t you hate it when you call a support number, and you have to first listen to a recording, then to some elevator music, that you are talking to some lady in India who you can’t understand and who cannot understand you, and then your are transferred three times to different support departments before your problem is solved?

    That’s what automation means, among other things.

    I trust that would not happen to me if I called BestRussianTranslator support.

    Report comment
    1. That reminds me: my phone is dead and I need to turn it on, in case someone will call me 🙂

      I believe there’s a good way to automate things and a bad way. The biggest problem companies have: they don’t understand that automation can also be human and personal when done right.

      Report comment
      1. Dmitri, you’re right about that. To me the difference between good and bad automation depends on only one thing: do the people that create a solution (to whatever they think they need to fix) listen carefully and with an open mind to the users of their product. This is also why creating a platform like Fluently is so hard: translation buyers and translators may have different opinions, or they may express them differently. You may have one loyal translation buyer, and you want to keep him happy even when he’s asking for features that may upset half of the translators using your product. Building a platform like Fluently is just a very complex thing because translation is not a commodity like most of the stuff that you can buy and sell at Amazon or eBay… Many have tried, many have failed. Still I believe at some point someone will do it right. That’s one of the reasons I always test new platforms.

        Report comment
        1. Thanks, Gert. Yes, I hope there will be a platform that will do it right. But I also think it will only happen at a smaller scale without major effect on the industry as a whole.

          If we look at modern technology companies and hot startups we’ll see that 99% of them don’t really care about professional human translators. Well, maybe they do care, but they realize that professional human translators are extremely hard to manage.

          This is why we’ll see more companies venturing into crowdsourced translation and machine translation/post-processing as a service. Because it’s much easier to control/make money out of the crowd and machines than professional human translators (who are being left behind at the moment by technology companies).

          Report comment
  4. Thank you for your post, Steve. I appreciate what you shared as I had not heard about Fluently before. I took the time to read through the article and found it very informative. I enjoyed Dmitry’s comments and learned a lot from you both.

    This post caused me to reflect on the number of people who choose to be translators under the believe that their interaction with other humans will be minimal. Many are shy and introvert, and probably enjoy being isolated from the world most of the time, but then reality hits, and soon many come to realize that overcoming their shyness is the first thing that needs to happen in order to succeed as a translator. We could be the best at what we do, but if we don’t know how to communicate with people and embrace human interaction we will not succeed as translators. I know this article is mostly referring to the role a project manager plays in connecting with the translator and his/her client, but I think this also applies to the role human interaction plays in our success as translators, our desire to connect and talk about our skills without feeling inadequate or not good enough, and our willingness to step out of our comfort zones to learn to connect better with other humans. I hope I make sense!

    Report comment

Leave a Reply

The Open Mic

Where translators share their stories and where clients find professional translators.

Find Translators OR Register as a translator