I will take this article with a bag of salt but I think we are very close to solving the human communication problem. Not because the machines are perfect but because the human, our benchmarks, are prone to mistakes and so a "close enough" machine would pass as a human just fine.