Having always been fascinated about communication (growing up, I was always keen on learning new languages and inventing signs and codes to ‘secretly’ communicate with my friends, without being understood by adults), I decided to experiment with communication barriers and the AI’s language problems.
It is common knowledge that Google translate is… well, not very good at translating. Similarly to what happens in the popular children’s game ‘Chinese Whispers’, I decided to put a text through several layers of Google translations (in this case, English, French, Spanish and Portuguese) and then back to its original language. (*) The result was somewhat hilarious.
I photoshopped this last image to depict the original text in Portuguese and the final result, back to Portuguese but having suffered the transformation of several layers of translation.
The reason automatic translations are usually bad lies in the fact that machines translate single words or groups of a few words and cannot understand the notion of context. To us, context is essential and it is often due to the context that we understand the meaning of words we’re not familiarised with or choose to use a specific synonym instead of other alternatives.
In the following example, I used the same original poem but different languages for each layer of translation (Japanese, Dutch, Finnish, Polish and Portuguese).
Here is the comparison between the original and the final result. All images will enlarged when clicked on.
(*) I could have, obviously, used any text to conduct these experiments. However, by using this particular poem, I was connecting my two main themes in Fluid Exploration together, which I thought would be an interesting and more exciting way to work with them.