One of the most famous tests on the field of artificial intelligence is called the “Turing test”, after the great British mathematician Alan Turing. It consists of a human tester having a conversation with an unseen entity, say through an Internet chat, and trying to decide whether the chat is taking place with a human being or a computer. If a program can consistently impersonate a human being, it is said to have passed the Turing test, and it is declared intelligent.
A similar kind of test can be devised with respect to translations made by machines. If the output of a machine translation cannot be distinguished from what a human translator would have done, we could say that the translation engine passed this restricted Turing test. As we have already mentioned, we are far from this situation right now.
But an entity that would pass this restricted Turing test with flying colors is the combination machine plus post-editor. It is the intervention of a human being that ensures the good quality of the output.
One strange feature of the Turing test is that a simple minded computer can be spotted immediately in some straight forward ways, for example by asking the product of two big numbers. A more sophisticated program would deliberately “lie” taking a very long time for the answer, and even giving a wrong answer to sound even more convincing.
Since the main advantage of using a machine is the speed with which the translation can be done, the equivalent way of “cheating” would be to hold back on the translation for a few hours, pretending that there is a human being trying to work it out bit by bit.
If you think that you can take advantage of this new technology, you can request a free quote at Translation Services.