Here is another solid look at the current online machine translation landscape. I was surprised to learn that the head of Google's translation effort "spends his days feeding hundreds of millions of words from parallel texts such as Arabic and English into the computer, using United Nations and European Union documents as primary sources." This must be a rather dull job, feeding text to a machine all day!

I also have to wonder about Osborne's analogy of translations to playing chess. If it is supposed to indicate that software can never overtake humans in this particular area, playing chess probably isn't the best analogy. I think that machine translation will never overtake humans where translation is required to communicate a message that has been wrought of high creativity, humor or idiom. However, when any content that was authored in a structured fashion, or initially authored with an eye to being translated (by humans or other means), we will eventually see machine translation that approaches 90% accuracy, and likely passes any Turing test applied to it.