That actually seems pretty impressive for machine translation. Also, you have to look at the rate of improvement in machine translation -- not so long ago, it w

That actually seems pretty impressive for machine translation. Also, you have to look at the rate of improvement in machine translation -- not so long ago, it was much, much worse. And Google is leading the pack here, and due to the rate of improvement so far (and this article!) we expect it to continue to improve, and it is exciting and positive to see this improvement, and surprised that we're being told how the improvement is performed! What do you propose as an alternative? Human translation of everything and no research in machine translation?