In Translation and proofreading
WILL AI STEAL OUR JOBS? (Fear of Machine Translation) - read the full article about machine translation, Translation and proofreading and from Freelanceverse on Qualified.One
Youtube Blogger
Is Machine Translation going to take over the industry? Coming up! Hello and welcome back to the Freelanceverse. It is finally here, I can finally give you the video that so many of you have been asking about. Is AI going to steal our jobs? Whats the future of translation? Whats going to happen with MT? Is it still worth getting into the industry? These questions and more Im going to address in this video. As I mentioned before, a lot of people have asked for this and Im gonna mention a lot of references in this video, thats why it took so long to get it out. I decided to just simply put a little reference number down here in the corner whenever Im referencing something, so its easier for you to follow. If you want to look up the source, you can just check the respective numbers, so I dont always have to cite the whole source in the in the video. Now to address the elephant in the room straight away: is AI going to steal our jobs in my opinion? Uh... yes, thats the short answer. Obviously in my opinion, its inevitable, but this is going to happen to every job, every industry at one point in my opinion. It is just a matter of time, its just a matter of when. According to Ray Kurzweil in fact, hes an inventor and a futurist, very interesting personality, hes also the the leading Chief of Engineering I think at Google at the moment, and his predictions tend to be very accurate, hes rarely wrong about future predictions. You can look him up in the description, its a very interesting guy and according to him, our human minds cant even comprehend yet whats about to come in terms of technological advances in the next 30 to 50 years. So if we are not even physically able to comprehend whats happening, so why would you worry about that. Thats kind of silly in my opinion, right? He even predicted that the state of singularity, meaning the point at which our human mind and technology becomes one and the same thing, will happen as early as 2045. Thats not a long time from now and thats scary if you ask me, but its also intriguing and it kind of puts worrying about MT into a whole other perspective. Theres much bigger things at stake, but lets take a step back here. Were getting a bit ahead of ourselves. I think for us translators its really important when we hear so much about Machine Translation and SMT, NMT all these words... its really important that we know what we are talking about and what is being talked about. So just briefly speaking: there are three types of Machine Translation over the years that came up. There is the Rule-Based Machine Translation, this was the first ever thing that was created. They started in the 1950s. Then there was or still is Statistical Machine Translation so called SMT. This came up in the 90s and has became then largely popular and largely used in different use cases in the early 2000s. And now the latest development is NMT, Neural Machine Translation which started in the early middle 2010s and became largely popular in 2016 when Google announced in a paper that they now switch all their Google Translate algorithms to Neural Machine Translation. Now within NMT there are different distinctions to be made between recurrent neural networks, convolutional neural networks... but this then really gets in-depth and its interesting I really like to follow this stuff, but if you are, then you can check the links in the description. Im not going to go into more in-depth into NMT. A quick comparison at this point between the two latest innovations SMT and NMT: SMT as I said stands for Statistical Machine Translation which obviously works on calculations right? It works based on probability of words and word phrases which then make up a model. For these probability predictions you need to have a large corpora of words and word phrases and while SMT is much cheaper than NMT, its obvious which drawbacks it has, because if you base language on calculations or numbers, it becomes too predictable and thats not what you want in creative writing essentially. What then happens is it becomes two word- centered, too centered on one specific word or one phrase and it cant take into account the larger context of the text and this is where NMT comes in. Neural Machine Translation is based on a neural networks, similar to the one in a human brain, in your brain. It has different nodes that communicate with each other. NMT then uses these nodes to create so-called word embeddings. They are like different clusters of words that have similar context, similar meaning or are just differently related, sometimes you dont know how these clusters are being formed and this way with all these cluster keys and word embeddings it can then take into account to a certain extent the context of a text. Thats why Machine Translation has become so good in the last couple of years. It can then also address the long-lasting dependency problem in MT, especially in German you often split verbs. You have one aspect of the verb at the beginning and one at the end and thats really hard for Machine Translation for obvious reasons right, because its technically the same word, but it has two parts in the same sentence and theres a long dependency on this verb. With these context cues, MT can take this into account and can translate these correctly. So NMT whats important is: it just takes into account past data to then perform a prediction for future data, basically what every kind of Machine Learning model does. Machine Learning, Big Data, AI this whole industry is a huge industry. The Machine Learning industry is set to reach 100 billion dollars market value in 2023 I think. Thats insane! So obviously theres a lot of drive a lot of push from governments, from big corporations behind it and thats why the progression is so extremely fast. Its mind-boggling, things you say today are old news in a few weeks already. The latest trend in Machine Translation are so called highly specialized domain specific Machine Translation models. Lets say you have a big corporation that works in the healthcare industry for example. They can have a custom Machine Translation engine made for the healthcare system and of course when youre really focused on this specific domain and you feed it with with millions of word data, this becomes very highly trained and very efficient, very useful, practical for this corporation. Another huge progression thats taking place is the so-called Adaptive MT. Adaptive in the sense that the engine adapts itself based on already confirmed segments. So if you have a segment above, then the engine immediately recognizes the context, the words from this segment, works it into the engine and the next segment could be changed based on the outcome of the previous segment. Of course thats very flexible and very useful in some use cases. Now, having said all this: where does the translator come in in this whole story? I dont make this video to scare you off or to not promote a career in the language industry, not at all. Then I would have the wrong YouTube channel, right? Thats the whole thing Im doing here, I want people to find their passion and what theyre good at and if its the language industry, I really want you to follow this dream and get into it. I definitely think there is space for both human translators and machine translators in the language industry. You know the amount of data created at the moment daily is... its literally unfathomable. I cant even... I think I read a stat in 2018, Im going to check the link and put the number here if you want to check it, this was 2018 and already back then 90% of all the data created by humanity combined had been created in the past two years, and that was three years ago. So imagine its probably even crazier now and imagine this incredible amount of data means that theres an incredible amount of data to be translated. And we should be happy for machines to be out there to help us with that, because theres so much crap and so much noise out there that just doesnt need the attention of a translator. Its perfectly fine if this is just translated with Machine Translation, because they wouldnt need... there wouldnt be the capabilities of even all human translators in the world combined to translate this data anyways. So what this then means that we can specialize on highly visible, creative, you know highly engaging content that really needs a special eye for for detail, consumer facing especially. And yet I see so many posts always on social media from people just taking a sentence out of context and maybe putting it into Google Translate and then it comes out wrong and then mocking MT and saying "this is why Machine Translation will never replace us". This is nonsense right... this is complete... it doesnt make sense the comparison. They obviously dont understand the bigger picture, because you cant compare a Google Translate tool thats free for everyone in the world to use with a highly trained domain specific Machine Translation engine. Thats not a comparison that makes sense and of course you can put a funny sentence into Google Translate and you can find it funny, but thats no proof at all that machines wont replace us. Anyways, I went a bit on a tangent here, thats not what my point was. My point with this is dont try to compete with machines, because you will lose. Its just a fact, we cant compete with machines. Computing power is just way way outperforming our brains, what you need to do is be complementary next to the machines right? And the valuable clients, the good clients they realize that its not about whether this is better, whether this is better, machine, human, it doesnt matter, instead the key is for a company to know in which use cases a translator is needed, in which use cases you can use machine translation, or in which use cases a combination of it makes sense. When this content is not consumer facing, when its just you know garbage noise that you want to make sense of, when its internal communication that never sees the light of day, sure use MT, I would probably do the same you know its much cheaper. When its highly visible though, when its consumer-facing, when it needs to be localized for a specific country or even a specific region within a country, you know hyperlocalization becomes more and more common now, or even with transcreation as well, so when youre actually not translating normally word for word, but you have maybe a joke or you have a lyric or whatever you have a marketing slogan that needs to be localized to a specific locale, you need a human translator. This is where we as translators need to focus on, this specific niche. We dont want to translate the garbage, no one pays for that. We want to be language specialists that add value to a company, plus there is the added aspect of Machine Translation Post Editing (MTPE) right? Thats a huge new domain that is coming up. Work for a lot of people consists now of post editing and there are still many colleagues that oppose to this practice, which you know is fine. I dont really understand why you would oppose to it. I always say I just charge it by the hour and I see it as a normal proofreading, because then I dont have any problems with it. I dont have any drawbacks as long as Im paid by the hour. Opposing it doesnt seem to make sense for me, because if you think about it, the translators who opposed the computer revolution back in the day, they didnt survive the switch right? So if you dont... if youre not able to adapt, if you dont go with the flow, thats just life right... it goes on. So theres definitely room for both, actually Tess Whitty wrote it perfectly in a blog post, Im going to read this quick she said: "Theres no need for MT and professional translators to remain at odds, theres plenty of room for both." I agree with that completely. What J just always recommend people, is to really establish yourself as a language expert, language specialist, language consulant, whatever you want to call it. Dont like give yourself the label of strictly translator, because thats not what you are right? You are a localization expert is also a good one and make sure you know at least a little bit about this technical aspect that I talked about before. Maybe just have give yourself a day or so to really familiarize yourself with this, because its important that we know whats going on in our industry right? And when it comes to the point that translation is really not needed at all anymore - as I said earlier in this video, our mind cant really comprehend yet whats going to happen, so theres no point in even arguing if thats possible or not - just when we get to this point and the translation industry is completely done, there will always be a language industry right? As long as there are humans, there will be a language industry, because thats our means of communication. And I have no doubt at all that I will find my place within this language industry, whatever happens. You can also look at it that way and that brings me kind of to my last point that I wanted to mention. Ive been thinking about this a lot actually lately... I think we often as humans worry way too much about like how can we optimize this, how can I optimize this outcome, if I do this and this... you know in 2016 when Google came out with this NMT paper, a huge uproar in the industry of course, and I was in my Masters in the Netherlands back then and I attended two separate lectures that kind of promised, well promised... predicted that translation will not be needed in the next five years. This could have scared me off, you know, this could have put me off from even following this career path at all, and this was exactly five years ago, and my business is still growing every year and theres still plenty of stuff going on in the translation industry. Its far from over, but thats not even my point, even if it was over, even if these five years and now it were over, I would still have done for five years of my life what I really enjoy, what I really want to do right? So sometimes I feel like... we can always change careers you know, its not that hard, if you have an education, if you have some kind of background to fall onto, its really not that hard to change careers. So yeah whenever people ask me like are you not scared of Machine Translation, you will lose your job... even if I lose my job, I will be fine you know, I will just do something else, but I know that in this world of language specialist, language expert, I will perfectly find my place. So maybe just dont worry too much about it you know, if this is your passion, if you love what youre doing, if you have the drive for it, if you are not scared to adapt, if youre flexible, if youre creative and if youre good at what you do, then just go for it really. There will always be a language industry, who knows whats going to happen in the future, no one can predict it, but I will be there and it would be cool if we were also there. There you go, thats my take on on future fears of AI etc. I hope this video makes sense, it was kind of hard to structure it all into one concise video. I hope it wasnt too dark I really wanted to show both sides, you know that there is a big hope actually, the language industry is growing, the translation industry is growing, globalization is becoming bigger which is good for us. Theres just endlessly more data to be localized... sorry my video cut out there just at the end, make sure you subscribe to the channel if you like the video and I see you next Monday with another video. See you bye bye!
Freelanceverse: WILL AI STEAL OUR JOBS? (Fear of Machine Translation) - Translation and proofreading