On opening the google translate app you find a variety of options. We can select to convert from a variety of languages to another in the list (which is more than 100). We can use handwriting to convert what we write. We can use the conversation service to convert what we talk into text and finally translate it to the language we want. And then there is camera. This is a very interesting option. As we open the camera it tries detecting a language and tries updating it to another selected language in real time.
It works really well when a picture is taken. But when a live translation is taking place the translation is very poor. For normal usage it is phenominal where in there is minimal lag between conversations while using the audio feature. This video shows the review of google translate by CNET.
This is can be bothersome as google translate can have trouble detecting and replacing the text in real time. There can be many factors that can cause this like, shaking the camera, lighting, processing speed of the device, etc. Only 37 of all the languages allow for real time camera detection. This feature is very attention grabbing as it detects the exact font and tries seamlessly changing it over the original image just like there was no change. There is an option to switch real time conversion on and off. In which case you can take a picture and select the detected words for translation.
The work on this translation although has jumped leaps and bounds still has more work to do for everyday usage in the real world. It is really good in converting website information, pdf ebooks, in the field of marketing by catering to a wide range of audience, and communication through voice. In the future there is no doubt that AR/VR devices will include this functionality. Under the influence and spread of AR/VR tourists will benefit a lot from this application as it will bridge the communication gap and travel anywhere once there is a language support for all the languages have been included. A person can learn to read common road directions and say phrases that are used in everyday conversations.
This capability can most efficiently used in the coming generations of AR Hololens/Magicleap on a wider field of view where different languages can be converted in the real world setting without lag. The user should be given the choice as then he would end up in a different city and not know the difference! But in all seriousness it can be really useful to have such a functionality that render the real world into an augmented world (which will be a matter of fact).
The main fact is that google translate is free and quick.
Google translate changes the input text as it is to the required language. It can be inaccurate and ungrammatical. When students use google chrome it can be a real burden for the teachers to correct them as sometimes the formed sentences can make no sense. Google translate works with a real database between two languages, by matching the exact words from one language to another. There can be issues where there are no words existing in that language for the one at hand. This can lead to generating some very ackward translations,i.e., the content can be lost in translation and it does not have a system to correct the grammatical errors that can be generated.
I think that humans should be given control over what they choose to see. Although a synthetic(augmented) world sounds cool in theory (and looks in experiments) there can be potential dangers. It can encourage people to perform or take part in activities that can disrupt the everyday life in the cities. For example, there has been a lot of Pokemon Go accidents and the technology is not even close to its final touchdown!
Realistically speaking in a professional scenario you would not use google translate and even if you do you will get it cross checked. Hence it has not reached a stage where it can completely replace a professional translater.
More recent developments in google show that since the wake of deep learning there has been a phase trying to recreate everything with deep learning. This is still in the nascent stages and it only supports 5 languages. It has a much better approach as compared to the procedural translation that the old model used, which was very "shallow". This article talks more about the new evolving google translate and its other features. Until there is a next big leap in deep learning this stagnant phase is going to be protracted for google translate.