It's the scale of data available for training to, there is a Google Research paper call 'The Unreasonable Effectiveness of Data' talking about how even in the n

It's the scale of data available for training to, there is a Google Research paper call 'The Unreasonable Effectiveness of Data' talking about how even in the noisy medium that is the internet there is a lot to be gained from that level of data. Also a paper from someone at Google on brute force paraphrase acquisition using billions of sentences combined with some relatively simple rules.