Behind the Scenes Part V

Please see samples here: https://www.youtube.com/playlist?list=PLO-QGEbwcTr2MdhbLPPGszMw8Rdc5J9aI

Behind these formal presentations of audio video recordings, there’s something very intriguing and interesting happening behind the scenes. In a very liberal sense, we are not too different from journalists or reporters who report on stories due to inspiration from those around them. What primarily drove us to begin and continue this project were the comments and feedback from interpreters and our colleagues. They have helped prompt and shape our actions, ultimately leading us to explore further, and continue our quest for answers on what the future will be like between AI and interpreters. Some interpreters are worried about not being able to make ends meet in the future, while others are asking why they should worry about AI. Will human language disappear? Will humans become like computers communicating with each other without having to verbally speak as certain individuals claim?

Whether we like it or not, AI will become more and more prevalent on public transportation, on social media, on our phones, in our daily life, and even in our industry. There seems to be a trend where humans make language simpler and friendlier for AI, so it becomes easier and more accurate for AI to translate or interpret. In this way, humans become a conduit for AI and AI becomes a useful tool for people worldwide to communicate with each other instantaneously. AI has evolved to become our translators and interpreters, and this application of AI has steadily become more popular over time. Ever since Google Translate was released, there have been increasingly new AI platforms such as AI video translation, AI spokesperson translation, and so on, that have surfaced. AI has been developed so quickly and accurately that it is just a matter of time until humans fully adopt AI. We tried our hand using one of the latest AI platforms to translate/interpret one of our office videos into Chinese. The results were impressive, but not without some imperfections. The AI voice did much better than the usual, generic robotic voice we typically hear, which most likely has to do with voice cloning.

In the face of rapid development of AI, does it mean that eventually we as interpreters will no longer be needed? How should we give guidance to the younger generation who aspire to become interpreters or translators? These are serious questions that make us really sit down and think and here is what we would like to share.

AI voices are serviceable, but they lack the beauty and liveliness that human voices have. That’s why human voices will always have a place in our world. This is what we set out to prove with the Interpreters and Voices series, and we think we have succeeded. In this blog, we are giving a conclusion about our thoughts on AI and human voices. Everyone knows AI voices are usually robotic and monotonous, but with voice cloning technology, AI could sound better and less robotic. This is why, even in the future, human voices will always play a role because they have that personal touch, which allows us to feel heard and assured. In the Interpreters and Voices series, 17 of our passionate interpreter colleagues have recorded themselves reading various blogs on AI nature and capabilities. Our colleagues have demonstrated how human voices are beautiful and lively compared to AI’s. It shows how big of a gap there is and how big of a gap there will always be between human voices and AI.

First, thanks to the participating interpreter Liling who introduces the concept of “aura” from a mechanical engineering point of view. She said AI-voice interpretations lack a key element – “Aura”.  Aura refers to a quality integral to an artwork that cannot be communicated through mechanical reproduction techniques and was used by Walter Benjamin in his influential 1936 essay “The Work of Art in the Age of Mechanical Reproduction”. Human-created artwork has its presence in time and space, its unique existence at the place where it happens to be, its ‘aura’.  Using this analogy, each interpreter recording has his/her unique interpretation in a unique space, time, and place, thereby creating an artwork with “aura”.

Second, while AI voices can only say exactly what they are programmed to say, interpreters are able to use their own individuality, smarts, and uniqueness to come up with clever phrasing that perfectly fits the situation, rather than just a word for word translation. Interpreters deliver words in a very exciting way that is palatable to the ear and also interesting to listen to. Each interpreter demonstrated their different interpretation of the script that we provided them. Some interpreters emphasized certain words. Others would, at times, speak at faster speeds. Some interpreters would ad-lib and sing certain parts while others would add a light laugh at a joke. Some interpreters go deeper with a dialogue style as if they are talking to each other, echoing each other, encouraging you to think further, or having a conversation with you. AI is not capable of this kind of responsiveness and communication. All the AI, regardless of which company develops them, lacks variation between them, making it easy to spot them almost right away even if it’s a clone of a human voice!

In the era of AI impacting every industry, including our translation and interpretation industry, these interpreters’ voices seem so wonderful and one of a kind. So we decided to create different collages of voices. We even have different colleagues work separately to create their own versions, and the results are stunning! Despite the colleagues making their own selections from the same pool of recordings, they arranged all the pieces together in different ways to convey their own take on the blog story. The entire project has been about human voices in range, richness, diversity and individuality.  It features different takes on the interpreters’ best individual moments, focuses on elevating each other to higher levels, and aggregates them to a beautiful, powerful collage as a whole. It serves as a reminder that we don’t want a world without human voices and also as a way to perhaps shed some light on the interpretation community. Quite a lot of work has been put into arrange everything for these purposes, but it’s a labor of love, and has been very enjoyable. We’re so excited to share it with everyone!

The video on Interpreters and Voices has a solo and two world versions. The solo version shows how powerful one person can sound while the world versions show how much excitement with many people from all over the world can generate. The world versions feature multiple interpreters from around the world, namely Asia, Europe, North and South Americas, and the Caribbeans. Each participating interpreter submitted their own unique and individual recording. Our colleagues combined them to the effect that it’s almost as if everyone is having a conversation with each other. It’s a truly beautiful collage of voices from around the world, all united for a forum discussion and you’ll feel like you’re right there with them!

It’s fairly easy to spot where different interpreters come from in the world, and it’s all thanks to their distinct styles. At times they are lighthearted, emphatic, or communicative that are embedded in their respective cultures. It’s truly a globally cultural feast! It’s also a showcase of the auras, emotions, and cultures prevailing in the world. Did you feel engaged listening to the interpreters? This collection of auras is something very difficult if not impossible for AI to mimic. We believe that as long as humans carry their own aura and pour it into their creations, AI will have a hard time getting a leg up over humans.

The video on Human Individuality, we’ve also created two versions: a trio and a quartet. The two versions feature three and four interpreters, respectively, and were also created by two different colleagues separately. The effect this makes is quite pronounced, and you feel as if you’re listening to two different pieces entirely.

For the short and sweet Thoughts about AI piece, we initially thought that with the way the content is structured, it’d be best for two interpreters to read it. But then what would happen if we added a third interpreter, someone who comes from another part of the world, instead of the duet of interpreters from the United States? The trio piece features wonderful chemistry between the three interpreters, and if you stay until the end, you’ll be rewarded with a surprise, which we are sure all listeners will enjoy! We have also made a solo version for those who might have been overstimulated by the different voices, and for those who might prefer just a single vocalist instead of an entire band. We enjoyed all of these so much, and absolutely recommend you check them out too! We’ve provided links to the audio and video series below for easy access!

Other than being fun, this project has also helped us see the bigger picture when dealing with translation and interpretation projects. Now we see our overall role very clearly and understand better what’s more important in the work we do. Therefore, we plan to show the series to younger generations, so they understand that translators and interpreters will always have work, and their value to this world will never change. The aura that naturally comes from being a human interpreter has and always will be something sought after. As long as we have energy and fun as humans and as interpreters, we will never have to worry about being replaced by AI. However, if we lose our energy or fun, we are doomed to surrender to AI superiority.

If you would like to encourage the next generation that there is a future for them, and they shouldn’t give up, we’d love to hear your advice to those who aspire to join our industry in the future. We will gather advice from all sources and present them in our next blog. We think that’s what we as translators and interpreters should aim for when we pass the torch on to the next generation in the face of a potentially AI dominated world! But we must always remember that interpretation is like art or music, or a fine-tuned performance, and that’s one thing area humans will remain dominant in for years to come!

Links to Audio Recordings:
1. Thoughts about AI (Solo Version): Posted
2. Thoughts about AI (Duet Version): Posted
3. Thoughts about AI (Trio Version): Posted
4. Diversity and Richness: Posted
5. Human Individuality (Trio Version): Posted
6. Human Individuality (Quartet Version): Posted
7. Interpreters and Voice (Solo Version): Posted
8. Interpreters and Voices (World Version 1): Posted
9. Interpreters and Voices (World Version 2): Posted
10. Human Aura: To Be Posted

Reference Video: Interpreters and Music

How does AI apply to the language industry? AI has always been a constant question mark in the back of our minds especially regarding automatic translations. Now that AI can do so much, is it time for translators to become worried?

Today, translators use machine translation programs such as Trados or MemoQ to give them a baseline to work from, so called post-editing machine translation (PEMT). PEMT can greatly improve translation times especially for longer translations. For translators, it has been recognized as a valid strategy and working method. However as we all know, the translations that machine translation programs produce aren’t perfect. They are often more than not flawed. But what happens if you begin using AI such as Chat GPT which is designed to produce fluent and natural sounding language? Would a translator have a more powerful tool? Or would you have a competitor?

Translators will find that the answer is the same as it always has, ever since machine translation was introduced into the world. A machine will most certainly not remove the need for a translator. Language is always so fluid, full of nuance and possibilities that it’s almost impossible for a machine to pick up on every last detail. On the flip side though, translators may find a new ally with AI as it can cut their work time, allowing them to work on more projects. For example, translators can enter questions into Chat GPT, maybe have it check for fluency or suggest different ways of phrasing things, or even come up with new ideas.

If you’re an interpreter, you probably have even less to worry about. Though AIs such as Google Assistant are popular for quick and easy phrases, it still runs into the same problems as it is pulling from Google Translate. Interpretation especially is a field where accuracy and nuance is important and a lot of that is easily lost with AI translations.

Overall, in the language field, whether the advancement of AI would replace the role of human translators and interpreters is nothing to be concerned about, at least not in our life time. If anything, we should be excited about it. As the technology advances, translators and interpreters may find a powerful ally that will help them advance their work efficiently.

First, there is Statistical Machine Translation, or SMT. The best example of statistical translation is Google Translate. SMTs such as Google Translate use statistical models that draw on a large amount of bilingual text that is provided to them. Through this, they try to find statistical matches for how many times something has been translated as something. For example, if the word “apple” has been translated as “manzana” in Spanish enough times, then that becomes the SMT’s pick for the word apple. It does this for every word in the source and target language before it finally outputs a translation. SMTs can be useful for translating basic words and phrases. However, for more complex sentences, SMTs are not a good pick because they do not factor in context. This means that you can end up with some very strange sentences that do not properly reflect what is being said in the source sentence.

Next, there is the Rule-Based Machine Translation, or RBMT. As the name implies, with RBMT, the machine is translating based on the rules of grammar that it is given. It goes over the source language to analyze its grammar and then looks through the target language to see what kind of grammar rules need to be followed when translating the sentence. After it is done, it translates the sentence according to the rules that it looked up. While this might sound as if there might be fewer problems than SMTs, RBMT translations still need a great deal of proofreading to ensure that there are no problems with the text.

Third, there is the Hybrid Machine Translation, or HMT. The HMT takes elements of RBMTs and SMTs and then uses a translation memory. The addition of a translation memory makes the overall quality of the translation higher than that of SMTs and RBMTs, but there are still problems with using a HMT. For one, it still needs extensive editing, meaning that human translators will be needed to review all the translations.

The fourth, and final type of translation is Neural Machine Translation, or NMT. This type of machine translation uses neural network models that are based on the human brain to make statistical models for translation. When they work, NMTs can provide some of the most accurate translations that would need little to no editing potentially. One of NMT examples is DeepL.

Overall machine translation can be used as a tool to expedite translation, but whether it can produce clear and accurate translation for high level discussions, complex concepts, nuances and context that is still in question and requires time to advance it. This once again proves that the value of professional human translators is not replaceable just yet!

This opinion has only been further enforced in the light of the current pandemic. With social distancing being encouraged, many businesses and individuals are moving away from face-to-face business and moving to the safety of digital solutions. With many people preferring in-person interpretation, and less interpreters willing to ignore safety conventions to fulfill these requests, there are some who would prefer to try machine translation.

In this day and age, there are sophisticated machine translation engines that can even almost do simultaneous interpretation. If the subject is within a specific, technical subject, the accuracy of the outputted translation goes up as well. Many would point to this as being the time for machine translation to rise in popularity and use.

Even so, businesses still tend to prefer using interpreters when they can. There is, and will most likely never be a true replacement to a human interpreter. In fact, interpreters were already still able to work remotely even before the pandemic. With Zoom and over-the-phone interpretation, interpreters are still able to work without too much of a change to the experience.

Of course, there are still some difficulties and annoyances with remote interpretation, but it is still the most reliable and preferred method for interpretation. The same applies for translation. Most translation has been done remotely, and is still preferred over machine translation for the same reason businesses prefer human interpreters. Though machine translation is a powerful tool, it is still not reliable or trusted enough to fully replace the human touch. During this pandemic, the work of translators and interpreters may be affected, but it is not because of a rise in machine translation.

As interpreters, we are always interested in the development of interpretation technology, including the possibility of “robot” interpreters. About a year ago, we wrote about the potential of interpretation software and Microsoft’s efforts in that area. Their software is meant to analyze your speech, translate it and synthesize an audio track of your own voice speaking in the target language. (However, the software will first need to study your voice for several hours. Also, it gets around one in eight words wrong.)

So are there robot interpreters in our future? The answer is complicated. Perhaps one day, but not any time soon.

Interpretation software faces many difficulties, in each of its component stages. The first step is speech recognition, an area that has seen great improvement over the last few years. Nowadays, speech recognition figures in our daily lives – when we make a call and are asked to speak to a robot, when we are driving and asked to give commands to the stereo system, etc. Due to its wide range of applications, speech recognition is the most mature of the technologies involved, but even it still has a long way to go before it becomes truly reliable, as we are well aware from our experiences.

Next, the words spoken (and hopefully understood correctly by the system) need to be translated. While machine translation has certainly come a long way, anyone who has tried to use Google or Bing Translate on a foreign language website can testify that what comes out is not always even legible, much less an accurate translation of the source text.

Finally, the software needs to generate a spoken version of its translation through text to speech synthesis. This is possible with today’s technology, but even there many irregularities remain. Intonation, emphasis and other vocal indicators still do not carry over particularly well into artificial voice generators, creating voices that sound, well, robotic.

Beyond these basic limitations of the software, there also remain concerns about the nuances of both translation and interpretation. Questions about homonyms, about words that have multiple meanings depending on context, and of course the issue of cultural differences. There are many situations where even the most sophisticated software will be unable to guess the subtle nuances a speaker intends. A human interpreter does not just translate the words, but also the subtext and hidden meanings within them, judging not only the speaker, but also the audience.

“Reading the audience” to judge their understanding and reception of the translation is a vital skill for an interpreter. It is also one that becomes even more important in smaller, more intimate settings outside of presentations and conference interpretation. In a business meeting, at a doctor’s appointment, even in court, the message needs to be tailored based on the reactions of the listener.

Can you image a robot, as they as now, doing medical interpretation or legal interpretation?

Monterey Language Services strives to provide the highest quality of translation and interpretation services. Please feel free to contact us for a quote.

Now, it won’t be long before a device is made that will enable audio to be translated to any language for display. In fact, a lot of transcribing services already provide the option to do this. However, a mobile device that could perform this feature would be newsworthy material.

Project Glass is still in its development stage, currently gathering patents on all its technology that’s being developed.  If this technology hits the market, the price is sure to be high. However, imagine a similar technology coming out for an affordable price. The means of technology are endless.

Picture and Info Source :
http://arstechnica.com/information-technology/2012/05/google-glasses-patent-hints-at-speech-to-text-display-for-deaf-users/