Understanding the point about understanding the point.

We are all familiar with the point that translation algorithms like Google Translate translate but do not understand what they are translating. In the race to design ever better translation algorithms, however, the point is often forgotten. These algorithms are indeed achieving great success and are a valuable tool for many people. But they can still be fooled, as Hofstadter convincingly shows in this article.

He neatly expresses the nub of the problem. Translation machines decode another language into the user’s own language. They substitute decryption for translation, using powerful statistical techniques. They possess, however, no contextual knowledge, let alone all the other features such as emotions that we humans have. As a result, they cannot cope with exceptional circumstances where the language in question contains ambiguities or non-public assumptions. In common parlance, they have no understanding. Hofstadter notes that we may create machines that possess these things, but that day remains a long way in the future.

Link to article: https://www.theatlantic.com/technology/archive/2018/01/the-shallowness-of-google-translate/551570/

You may also like to browse other Language posts: https://www.thesentientrobot.com/category/language/