Category Archives: Machine Translation

‘Leading linguist’ says mediocre communication is our future

Posted on by

According to a recent article in the Telegraph, ’leading linguist’ Nicolas Ostler has stated that English as a lingua franca will soon die out, to be replaced not by a new language of trade but rather by machine translation. I have no qualms with the premise that English will eventually cease to be the ‘global language.’ Languages, like people and civilizations themselves, eventually pass away. However, as I posted just a short time ago, I don’t see machine translation truly becoming ‘translation,’ mainly because without a true brain behind the wheel, it is quite literally impossible to accurately translate any sort of content. Of course, many people are pushing for the machine translation revolution, Google included. Unfortunately, however, without true artificial intelligence powering your machine translation, all you end up with is garbage data that gives you a vague idea of what the original content actually said, usually with mixed up negatives and literal translations of figurative expressions that can lead to serious consequences when used for actual communication purposes. Google Translate is the state of the art in the field and even theyadmit both that it’s just about reached its limit and should only be used for gisting, not actual communication. In other words, the ‘future’ of inter-language communication according to Ostler and others is already here and that future is mediocre at best, dangerous at worst. To give you a taste of this ‘glorious’ future, I’ve taken the liberty of translating this article into Japanese and back again using Google and posted the result below.

 

According to a recent article in the Telegraph, hostler Nicholas scholars “major languages” is not in the new language of trade English as a common language is not to soon be replaced by machine translation rather, there is to go extinct said that the funnel. English is the language I no pangs of conscience with the premise. “Global language”, as people and civilization itself, and eventually passes away to cease to be final. But I have posted a little while ago, I if there is no true brain behind the wheel, the main reason is literally impossible to convert any sort of content precisely, the machine translation it does not appear in the ‘translation’ really. Of course, it contains a lot of people are promoting the revolution of machine translation, and Google. However, without the need for true artificial intelligence unfortunately, to supply power to the machine translation of your After all, it is garbage data, or a literal of figurative expression may lead to negative is usually mixed when used for the purposes of the actual communication serious consequences give you an idea of what vague translation, the original content of what you said actually. Google Translate, and even a state-of-the-art field, they both admit that should be used to create only the gist of it is not only the actual communication, have about reached the limit. In other words, According to the hostler, “future” of inter-language communication, where its future is a commonplace in critical condition, the best of the worst already. In order to give you a taste of the future “of glory” this to you, I will use again and Japan, Google has posted the following results, a letter of translating this article.

If true machine translation becomes a reality, everybody (not just translators) is screwed

Posted on by

The only way I see true, usable-for-any-situation MT becoming a reality would be with the invention of true AI. Translating meaning necessitates understanding and understanding necessitates consciousness. Especially with context – in Japanese, whole passages can go by without the subject being explicitly stated, which works fine and dandy for Japanese but makes it impossible to accurately translate it into English on a sentence-by-sentence basis.

However, even if we ever managed to create a true, human level (or beyond) AI, would we use it just to translate some random financial documents? Assuming that a true AI can be created, I see two possibilities as to its nature: either each AI is unique and has to be ‘grown’ through experience like a living person or you can make unlimited copies of the AI and run it on any computer that has the CPU muscle to power it like any other piece of software. If it ends up being the first possibility, it will be far too precious to use for ‘mere’ translation and far too expensive to make it profitable even if you did. If it ends up being the second possibility, that AI plus some general-purpose robot bodies (like the Honda Asimo already in existence) means that virtually every person on the planet will be out of a job and we’ll either enter a sci-fi paradise where people only work for pleasure or most of the world’s population will instantly be out of a job and revert to a literal hunter-gatherer existence to survive on the outskirts of Detroit et. al. while the handful of CEOs that run the robot and AI companies live a cushy existence until their eventual overthrow by their cybernetic slaves.

Even if an AI is doing the translating, you’d need to equip it with many of the same tools a human translator needs, like internet access to research new terms and confirm terminology use (as language is constantly evolving, especially in technical fields) and even a phone line so it could confirm, for example, names of businesses that have no website, which is a big concern in Japanese and other languages that use a writing system not descended from Latin. Or you could just have it email those names to you so you can confirm them. Not deal-breakers, to be sure, but that AI translator starts to have the same concerns and drawbacks that a human translator has, though presumably it would work like a slave for free once you bought or rented it.

My ultimate concern, though, would be whether you could truly trust this AI, especially with sensitive information. What if it got hacked? Your company could lose all of its trade secrets in one blow because everything was being funneled through the AI for translation and all of that data was sitting in the AIs files for future reference. Further, how can you be absolutely sure that it’s translating everything correctly? Of course, human translators make mistakes as well, but how do you check the machine’s translations? With a human translator? With another, but different AI? And then there’s the fact that it’s just not a human intelligence -it’s not using a human brain, evolved over countless generation to have specific features that operate within a fairly definite range even between different people. It would be using software and a computer, or maybe a specialized device designed to facilitate AI, to operate. Either way, that’s not a human brain and it’s not going to think exactly like a human. If you give it its freedom and allow it to build trust like any living person – by working and not screwing up and holding it accountable for its action when it does – that’s one thing. But trusting what is effectively an alien slave is probably not something a lot of companies would want to do when faced with the reality.

But I should stop now that I’m neck-deep in science fiction. I think Steve Vitek said it quite eloquently, however, on his blog when he asked “Do you think that it is possible to create software, similar to machine translation software, that would write steamy romance novels that women would actually be buying and reading?”