Man vs. Machine: The Direction of Machine Translation and Questions on Its Implications


Machine translation (MT) is not the new kid on the block. It dates back to to about ~1950. But though translators have (somewhat hopelessly) been arguing about the pros and cons of MT for quite some time, it wasn’t until recently (when literally millions were poured into MT by Microsoft and other IT giants) that the general public also joined in the debate and marketing turned MT into the greatest human invention after the wheel.

Meanwhile, for quite some time, linguists have been observing a trend: more and more agencies are selling human edited MT to end clients, which means more and more linguists are shifting from translators to “MT post-editors” or other colorful terms used to describe them. The human vs. machine debate is fascinating from a linguistic point of view; and humans win every single time. But money speaks louder than words and MT saves millions; therefore, despite well-founded warnings from the bleeding hearts of the translation world, the market continues to shift toward MT with human post-editors. However, the fact that we can’t stop this train, doesn’t necessarily mean we have to hop aboard without at least asking ourselves where it’s going.



Confidentiality suits cost the world millions every year, especially in the developed world. The potential legal implications of MT in regards to confidentiality have been analyzed in several papers and blogs. However, Matthew Blake summarizes it best in layman’s terms: “According to many legal experts, the actual sensitivity of confidential information and the ongoing efforts to keep it undisclosed are necessary to keeping information confidential. If the owner of the confidential information is reckless with the information, is it truly confidential?” What I would add to Matthew’s question is what happens when the person being reckless with the information is the translator or agency entrusted with it?



When using MT, such as Google Translate or similar technologies, you may find things like this in your service agreement:

“When you upload, submit, store, send or receive content to or through our Services, you give Google (and those we work with) a worldwide license to use, host, store, reproduce, modify, create derivative works (such as those resulting from translations, adaptations or other changes we make so that your content works better with our Services), communicate, publish, publicly perform, publicly display and distribute such content.”

This obviously does not play well with intellectual property rights; and because this is a huge issue, I will address it in a separate post.



MT raises an array of ethical questions from several different perspectives.

1) Affecting clients: Are clients being told that MT is used? If so, are clients aware of the confidentiality issue? Are they aware of the intellectual property issue? Is the rate they are paying consistent with the translation quality they are receiving? Do they fully comprehend the implications of machine vs. human translation? In other words, are clients giving informed consent to the use of MT or are key potential issues being withheld or concealed behind pricing strategies?

2) Affecting agencies: Who is ultimately responsible for any breaches to confidentiality or intellectual property? Can agencies keep their client’s information safe? What does that duty imply and to what extent are agencies expected to take measures to ensure such rights? How far can agencies go to control their translators? What is a legitimate business practice in the MT framework and what borders on abusive? What are the implications of having inexperienced or unprofessional translators post-edit what is already low quality MT language?

3) Affecting translators: What about translators’ intellectual property rights? How much are translators willing to waive? Are they actually required to do so? How does the new MT-based business model affect their income? What happens to translators in developing parts of the world that are already more vulnerable to abusive practices and have limited access to necessary tools for competing in the current market?


These are just some questions off the top of my head that I believe we should probably think about before submitting anything to MT. Needless to say, these are not easy questions and it may take us some time to find the answers. In the meantime, I wonder what my readers think…

12 thoughts on “Man vs. Machine: The Direction of Machine Translation and Questions on Its Implications

  1. Very good points.

    There are also concerns about the psychological damages of the exposure to poor MpT language could cause, as well as damages caused by the mind-numbing process itself.

    Just a couple of comments:

    I wouldn’t necessarily say that MT saves millions. Trained MpT (Machine pseudo-Translation) is very expensive to setup and maintain, or in other words, costs a lot of money. Using generic engines (such as Google Translate and Microsoft Translator) may save some money in the short-term, but their true cost (including damages to reputation, loss of business, etc.) may be long-term.
    I also respectfully disagree with the statement “the market continues to shift toward MT with human post-editors”. In my view, there is no single-translation market, but many markets. The segments or sub-markets moving to the MpT clean-up model are mostly those in which poor (cheap) human translation with human post-editors is the standards for years now. The move for MpT is not due to its professional benefits, but for the sole purpose of replacing the unreliable cheap translators, while maintaining or increasing the profit margin of the broker in the process.


    • Thank you very much for your comment, Shai. You raise some very interesting points.

      Although I share your sentiment that MT has its costs and that not all agencies or large companies are immediately hoping on the MT train, of the $37 billion that reportedly went into translation in 2013, $4 billion went directly into the top 50 LSPs. I checked out about 25 of these companies and they all advertised MT on their sites as a cost-saving part of their process. If we part from the assumption that these companies are not in the business of losing money, I think it’s safe to say the cost of MT (though probably high) has not outweighed savings for them just yet. In addition, we’re currently being bombarded with articles for hire on major newspapers and other media (including recently The Economist), where industry leaders are encouraging the use of MT and ignoring the big questions and issues it raises.

      Personally, I don’t think MT has much to offer, but I think we can’t deny that a very large market segment is shifting in that direction. Since Google and Microsoft recently launched their new MT toys, the public is being bombarded with pro MT propaganda and MT has become an easy marketing strategy for capturing linguistically unaware, price-oriented clients. Although many market segments may stay clear of this trend, I think we have enough empirical evidence to think the trend is there and to start discussing how to raise awareness in the opposite direction, i.e. how do we get people to understand MT is not all that it’s cracked up to be?


      • The top LSP (what a silly and misleading name) list by the common nonsense advisory is not exhaustive. Let’s just say that they look only into what they are given access to and have interest in promoting.
        Real MpT costs a lot of money. Most optimistic and realistic estimates talk about a 10%-15% cost-savings when translating very large volumes of content. This is not that big of a saving when you consider the quality hit, but it can save money on things that otherwise probably wouldn’t have been translated or were sent to translated by the cheapest translators available; i.e. for gisting purposes. The million DIY offerings out there are the real hoax here.
        Now, these companies are not in the habit of loosing money, quite the contrary, but it doesn’t mean their clients save money, nor that their entire revenue is based on MpT. Furthermore, all these reports don’t give any insight into which documents are being MpTed, i.e. short shelf-life or very low priority documents, sometimes just for content discovery or gisting purposes.

        Some industry “thought leaders” and “influencers” are actually not reliable. Some had sold out to companies/lobbies and now promote their interests.
        The PR propaganda efforts are indeed a problem, but mainly because there are only few who fight to tell our narrative, and their resources are dwarfed in comparison. This is not surprising considering most translation associations are ruled by agencies, the ATA included (re: the Economist).

        People will learn with time about the damages of MpT. MpT is a professional tool, for some. It is a very nice and useful aid for private users for gisting and discovery purposes. The hoax lies in the middle, in the way some profiteers are abusing the technology for things it is not designed for. The profiteering commoditizers are riding on the wave of obsession with technology that is sweeping the world, and way over-promise. Those market segments are the same segments that translation professionals had nothing to do with for years now.

        We should make our voice heard, and demand our associations to invest in PR (or leave them and turn to associations like IAPTI), I completely agree with that. The MpT technology is interesting, but there is a difference between the technology and how it is commercially used/abused. This will not be the first time a craze will eventually clam down, just to leave humanity to deal with the consequences. The most recent, and quite similar one, was the multitaksing craze of about a decade ago. Today it not even remotely a thing, actually, people are warned against it. Just seven years ago or so there was an entire industry around that silly notion. The resulting ADHD, however, is still quite a thing, and probably continue to be for some quite time now.


  2. I think there are a lot more assumptions about what I am saying than acknowledgments of mutual understanding in your reply. Is it possible we agree more than you think? It seems to me, the only thing we disagree on is whether or not LSPs (or whatever you prefer to call them) are saving money.

    1) I’m not basing my assessment on the CSA report alone. Give me a little credit! I actually took the time to check out about 25% of these companies (including their annual reports), how they advertise MT on their sites, and what strategies they are using to sell MT to their end clients. I’m still looking into the rest, but it takes time, and -unlike CSA- I’m just one person!

    2) I’m not saying MT actually IS a big money saver for end clients; what I am saying is that it is safe to assume that it is a big money saver for LSPs. These companies have information, which we don’t have. The information that we do have are their annual reports, and we know they make millions. We also know that they are actively advertising and using MT. We know that their profit margins have increased in recent years, meaning their business models are increasingly effective. We know that large companies base their business models on cost-benefit analyses. We know that the rational business choice when a cost-benefit analysis is negative is to drop the model. They are not dropping the model. Therefore, it logically follows that the model is working for them for now. None of this is an argument in favor of MT from the end client’s point of view or even my point of view, for that matter! It’s also not an argument in favor of long-term use of MT. It’s just a logical assessment of a current state of affairs. I don’t like MT any more than you do, but what is the point in denying that LSPs are not really saving money off of MT when all the evidence indicates that the model is currently working for them?

    3) Nobody on the opposite end is taking the time to translate “quality hits” or the like into financial terms. Until someone does that math, we’re just speculating. We assume that MT costs more than it saves, but we don’t have the necessary information to back that up. All we know, is that the big money makers are switching to MT to keep making even more money. Doesn’t that tell you something about our assumption?

    4) When I mentioned industry leaders, the intention was simply to reply to your statement about how there is no trend. My argument is that if there is no trend, then how do you explain all the money and PR or the number of companies literally boasting about their use of MT?

    5) I’m all up for making our voices heard. I would kindly invite you to read through more posts on this blog, there is a big fat ideological message behind almost every single one: don’t let yourself be exploited, defend your rights, strive for quality, and don’t buy into the hype. However, as much as I care about getting our message across, I would not “demand” that non-profit organizations do anything in particular. I would, instead, encourage them to voice their opinion and encourage members to help spread the word. I would have fundraisers for PR campaigns. I would help out and contribute to raising awareness, etc. But “demanding” from a non-profit just doesn’t seem right. Their resources are limited and they do what they can. Paraphrasing JFK, “ask not what your association can do for you, ask what you can do for your association.”


  3. Hi Paula,

    It seems like we have a little misunderstanding on our hands. I agree with you on many accounts, except for one thing (I’ll get to this shortly), and certainly not criticizing you.

    2) For some brokers MpT is a way to maintain or increase their margins. No doubt about it. MpT is not adopted by clients due to mertis, it is pushed upon them by brokers seeking to make more money. If you have a little time, please watch this video that “reveals” the mindset and motives behind MpT. Here is a video. from TAUS this time, about how to pricing MT Post- editing from the broker’s percpective. Note the mention of discounts, how to convince translators to accept these discounts, the idea of piggybacking on the unethical “established” ‘CAT-discounts’ scheme that the poor souls working with these brokers are accustomed to, and the complete absence of any mention of success collaboration case-studies, the mental hazards of PEMpT, fair compensation for one’s work, and ethics. And here is an article discussing the psychological aspects of how to “convince” people to post-edit.
    So, yes. You are right that from some brokers, and probably to varying degrees, MpT is way to maintain or increase their margins.

    3) Those who use MpT this way are not concerned with quality. They didn’t provide quality even before they turned to MpT. The business model of these brokers is all about driving volumes. They are focused on working with corporations that spit out a lot content. This is the where I respectfully disagree with you slightly. This is only one (or few) market segments, with a very specific way of working. There are also other market segments, approached, and realities. Focusing on this market segment as an indicator is not very different from focusing on the Fast Food industry and deducing from it about the state of the food industry. Not the best analogy perhaps, but the best I can come up with now.

    4) There is a trend, and it is in the best interest of some to make it seem bigger and more universal applicable than it really is, or can be. I never denied there is a trend, I just commented that it is not global, because there isn’t only one single translation market.

    5) Well, the associations might be non-profit, but the members pay membership fees to have their interests and profession promoted, and therefore I think it is well within their right to demand to be represented. If the associations fail at that, I don’t see the point in paying them at all. I understand limited resources, and if that was the problem but otherwise the available resources would have been spent wisely, I wouldn’t have a problem. But it is not really the case.because agencies — and their money — are ruling the associations and using their membership to leverage and exert influence, and as a platform that gives credibility to their self-serving propaganda, even in the eyes of buyers. This is wrong and unethical in so many ways, but yes I know, this is how the world and politics work.
    If MpT will eventually “take over” the world, it will not be based on merit, just the result of snake-oil marketing, and the naivety of people believing technology can offer them a magical solution.

    So if to conclude, we agree on almost everything, but perhaps differ a little in terms of overall perspective. And I just want to reiterate that I just write my opinion, and not criticizing you or anyone else in any way, shape, or form.

    Liked by 1 person

  4. First, of all, Shai, thank you for taking the time to comment in the first place and then engage in this discussion with me. I’ve seen you in other forums, and I always find your point of view to be very enriching. That said, I did not interpret your comments as criticism in the negative sense; on the contrary, I found them to be very constructive and wish more readers would engage in productive discussions.

    I’m glad that we agree on more than we initially thought, it just makes it so much simpler and more rewarding to discuss the points where we don’t agree. Some thoughts on your last comment:

    1) Propaganda: In my Matrix series on this blog I discuss how translators are sold the notion that “rates are dropping” to get them to agree to low rates, when in actuality, more and more money goes into translation each year; and there is no need to work for peanuts. It had not occurred to me to look at MT as an “agent” of the Matrix. I will look into all these resources you generously provided and think about that a bit more. So thanks for all the links and materials!

    2) No such thing as a single segment: I agree. Market economies can’t work without market segments. I have also addressed that in this blog in my posts on rates, where the conclusion is basically to stay away from the commodities segment and concentrate on specialty segments to make more money as a translator.

    3) Trend: In retrospect, I think you’re right; “global” may be too strong a word to describe the trend. I should probably choose my words more wisely next time.

    4) Associations: I recently joined the leadership council of one of the divisions of one such association and what I’ve seen so far are good people, most of whom are also translators, doing their best with what they have and who genuinely care about the issues that face our profession. There are exceptions, but I can’t say exceptions speak for the whole. I can see how one would be tempted to think that if they are supporting an organization with their membership, then the organization should strive to represent their interests. That is absolutely logical. However, I see a problem with that logic and it’s that, because translation is so segmented, associations represent individuals with radically different interests. “Which to tackle and how” is the million-dollar question. Finding a middle ground between member interests can be rough, which is why I believe in active leading and initially objected your use of the word “demand”. I believe in speaking up, taking sides when necessary and getting involved. But associations are only as strong as their weakest links, so we need members to be strong, outspoken and pitch in. Perhaps that’s also a way of “demanding” things from our organization.


    • Thank you for your reply, Paula.

      I agree with all that you said. Just a couple of remarks to round things up:
      1) Propaganda: As I see it, the big brokers (aided by some technology developers; there are commercial affiliations at play) are engaged in a social engineering effort with the goal of creating the appearance they–and only they–hold the keys and insights into the translation market, and therefore it is in the best interest of translation buyers and translators alike to listen to them and do what they told, essentially locking them down in believing they are dependent on the brokers, The example you gave about the “rates are dropping” notion is an excellent one that demonstrates just that.
      In this scenario, MpT, other technologies, and other factors are just tools, or agent of the matrix as you so cleverly describe them. The danger does not stem from the technology itself, which may or may not be useful as an aid to some degree , it is all about its abuse by people. MpT is a tool, and today we have too many irresponsible people running around wielding tools they don’t know or understand, in an almost desperate attempt to pocket as much money as they can before the hoax will be reveled.

      4) No doubt there are good people in the associations. However, good as they may be, they don’t necessarily know how to rebut or counter many of the PR efforts made by the commoditizers, who pretty much use the services of professionals. The road to hell…
      What you said about associations having to cater to many interests is the crux of the issue. An association should represent a relatively homogenous group. Of course, there are always some diversity, and this is a good thing, but only as long as it is in the form of poles or extremes of the same interest group, or to a lesser degree different profile groups with with a symbiotic interests. For example, an association that helps guiding less experience members how to go about their career is a good thing. On the other hand, an association the gives stage to those who work against the interest of other members, is betraying the trust of at least some members.

      The profession has survived adversities before, and will survive this one as well.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s