Posted by: patenttranslator | August 4, 2019

The Machine Translation Time Bomb

The most valuable asset on earth is no longer gold or oil. That may have been the case in the past, but times have changed. The most valuable asset in our world is now data. Including your data, among all kinds of other data. Just look at the stock market – the wealthiest and most powerful companies in our world at the end of the second decade of the twenty first century are tech companies selling mostly data and data-related services, companies like Google, Facebook and Amazon, to anybody willing to pay. 

Not enough is being done to protect this data … but that is not really all that surprising, or what I want to write about in my blog post today. There is simply too much data out there and too many hackers are trying to steal it to get fabulously rich. The smart ones may eventually succeed, the not-so-smart ones may end up in prison.

What is surprising to me is the cavalier attitude of some companies, including large, very profitable corporations, to how this data is created.

For example, software designed and used to safely navigate airplanes through night skies when hundreds of passengers on the plane are trying to sleep, mostly in vain, is one example of very valuable data. And because this software data is so valuable, companies that create this software to use it for their planes should be very, very careful about how and by whom this software is compiled.

But when profit is king, corners are often cut in the most inappropriate places in a manner that may ultimately lead to disastrous results. Because writing and designing this kind of data is expensive, Boeing outsourced writing of the software code to subcontractors relying on software engineers located in countries lacking a deep background in aerospace design, often in India, who were being paid $9 per hour.

I tried to find out how many people died in the two recent crashes of Boeing planes, but after about 10 minutes of furious clicking, I gave up. I did find a number of articles making a general mention of these crashes, but without the numbers, and each of the articles started by stating that “Air travel has never been as safe as it is now”.

It is impossible to find this kind of apparently top-secret information from Google or from another search engine. Clearly, since Boeing executives failed to exercise due diligence when it came to hiring qualified and experienced programmers, although I don’t know how it is being done, they are now exercising due diligence when it comes to removing the relevant information about the number of people who died as a result of criminal negligence of the company from the Web. It may be only a matter of time before the link I placed in the paragraph above disappears from the Web too. How much money will Boeing ultimately lose because it was cutting corners by outsourcing the compiling of critical data to the cheapest source available? Nobody knows yet.

What does all this has to do with what I call the time bomb of machine translation?

A lot, in my opinion. The use of machine translation is celebrated by the so-called translation industry as a new, innovative way to make translation available for a fraction of the cost of actual human translation to thousands of customers, in particular corporate customers, by using machine translation, which is virtually free, and outsourcing the editing of the machine-generated product to cheap subcontractors who are paid about 1 cent per word. These subcontractors would by necessity have to live in third world countries where 9 US$ an hour is a pretty good wage …. that is if you can proofread 900 words per hour, looking at the original language that you presumably know well, and the language of the translation, which you presumably know as well too, and trying to fix the machine-generated output, which according to the “translation industry” is very reliable and needs only a little going over with a fine-tooth comb.

The industry keeps publishing studies about how accurate and reliable the machine translations are. I don’t remember the actual statistics that the industry is using, although I glanced at several of them. But I do remember that the alleged accuracy is very high, 90 percent or something like that.

Which is a completely bogus, purely propagandistic number, of course. If you translate a long sentence ending with “in our expert opinion, the real estate property is not worth 2.75 million US dollars,” for instance from German to English, as “in our expert opinion, the real estate property is worth 2.75 million US dollars”, the translation is 97% correct when one uses the laughable statistics of the industry. The only problem with the machine translation is that the software missed one word, namely the world “nicht” (not), which as I and many translators know is sometime missed by machine translation software, probably because it is often located at the end of a long German sentence combobulated to an extent that makes it difficult even for human brain to discombobulate it, let alone for software.

An that is where the machine translation is located, in plain sight, although nobody seems to see it. The “translation industry” certainly never mentions it! A company relying on a machine-translated text, even if the were to be later “post-edited” by cheap post-editors whose eyes have to keep jumping between the original language, (which is a language that they presumable know well) and the machine-text, (which is in a language that is presumably native or close to native to these rushed post-editors), if they want to make at least $9 per hour, would stand to lose mucho, mucho dinero as bearded bandits like to say in American westerns.

What is the percentage of translations generated by the “translation industry” that is based on machine translation output, glanced over later by human “post editors?” Nobody knows this number either. The “translation industry” does not mention it to its clients, does it? But I think that in a world where the maximum profit is king, it is probably a very high number.

So I repeat, that is where the time bomb of machine translation is hidden. Fortunately, the bomb can be fairly safely defused by companies that do not trust the “translation industry” propaganda, avoid the “translation industry” and work only with certifiable humans, because only humans can generate data that is based on human cognition rather than data that is based on an algorithm.

Because as we all know, algorithms can sometime go amok in a way that even a well-functioning human brain is later unable to detect.

Advertisements

Responses

  1. Hi dear:
    Besides being a translator, I have worked and still do in cybersecurity.
    The example of Boeing you gave is absolutely true. The consequences of this are so bad that they actually did pay Google and other “Crisis Management” companies to remove from the Net any references to these facts. They also generated an enormous amount of references on positive statements that any Google search would offer if you try to find out about it.
    Yes, they CAN do it, they have the castle and the keys.
    It is also true, what you say about Neural or Machine Translation.
    All of this is fake news and they do not work very well.
    Now what is really scary is what the data companies do with your personal information, this is really scary: they help a clown being elected as POTUS, produce a yes to Brexit, help a Nazi be elected in Brazil and so on.
    We see only the visible part of the Iceberg (it is very big) and very much out of control.
    The EU and France, in particular, are both trying to have the GAFA pay some taxes somewhere in Europe not taking advantages of the tax loopholes.
    The problem is that the fight is not fair: a company such as Google of FB is more powerful than the European Commission. They will never pay taxes, or even if they do, it does not matter. Apple had a fine of 13 billion Euros to pay for Ireland. Well, guess what, Ireland told Apple that the fine was simply erased. Apple had a talk with the Irish Government.
    Since the member states of the EU can do as they please, they did that.
    Amazon sells stuff all over Europe and they charge you the VAT. Their headquarter is located in Luxemburg. The former prime minister of Luxemburg did a nice favor to Amazon, within the EU the VAT is compensated, that is the money they charge you in country A is paid in country B where the headquarter is located. So it should go into Luxemburg. But they did a special law for Amazon saying they need not pay the VAT at all. How do you like that?
    the VAT varies a bit in the EU but it is around 20%, which goes as pure profit to Amazon’s bottom line.
    Well and this very unknown law passed in the US in march 2019, saying that any cloud service provider that has at least one subsidiary in the US, is obliged to deliver all the info requested by any AG in the US? Nice tool for American companies and US Intelligence services.
    The worst company is FB. There the product is you and your personal information. And they sell it and they don’t care about privacy and GDPR and other crap.
    Zuckerberg should be in jail, but once again, you cannot put in jail somebody with the power of this guy. Let us imagine that he says “you want me to go to trial, OK, I cut FB until the trial is over…”.
    No luck there.
    Be well and prosper
    Ivan

    Liked by 2 people

  2. Actually, I found the numbers with 3 short clicks for “Boeing 737 crashes”–157 people were killed in the 3/10/19 crash of Ethiopian Airlines, and 189 people were killed in the 10/29/18 crash of Lion air.
    But I agree anyway with your premise–because Google seems to hide items 1 day, while the next day, they are very easy to find – with the same search criteria.
    An amusing note–searching for a local classical music program, I clicked on a program with a Russian performer–and the site was blocked (I could unblock it). When I clicked on another program in the same series–no block (here in the US), of course.
    MT is the bane of a translator’s existence. But one needs to remember that MT has been built up from quality human translations. I shudder at what will happen when MT builds up from MT. As you mention, good customers know the difference between human and MT …
    Another amusing note. I received an email from a hopeful vendor (I use this word because it is a company rather than an individual). They promised high-quality SLOW translations because quality translations take time.

    I enjoy your articles Steve–good luck in your new abode and in semi-retirement.

    Alice Berglund

    Liked by 1 person

  3. Thanks you for your comment, Alice. Unlike myself, you clearly are an investigative genius. Your next mission, should you choose to accept it, would be to find Trump’s missing tax returns.

    And thanks for your wishes of good luck in my semi-retirement as well. After I moved from a wastefully large American house that had 350 square meters (3,800 square feet) to a small Czech studio of 30 square meters (about 350 square feet, I think) where I have been living for 10 months now, taking pride in my newly found minimalism, I was overcome by claustrophobia and I will soon be moving to a new place that is twice as big and costs only 1,500 Czech crowns (65 US dollars) more.

    In addition to a more abundant size (for one person), I will have wonderful views of green fields and forests from the living living room, bedroom and balcony of fifth floor, with a smallish mountain called Klet in the background, where I used to go hiking and discussing Seneca and other sages with my friends when I was a teenager a long time ago. I remember that the last sentence we discussed when I was seventeen was “Natura duce, errare nullo pacto potest” (You can never go wrong when nature is your guide).”

    Hope you’ll wish me luck in my new abode too.

    Like


Leave a Reply

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

WordPress.com Logo

You are commenting using your WordPress.com 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

Categories

%d bloggers like this: