In light of the recent monopoly hearings in the Senate, Mr. Schmidt is clearly trying to persuade Senators that Google is weak and vulnerable. But Google published findings that show that in natural language processing, it's really the quantity [1] that determines quality. Note in particular the discussion of "stupid backoff" (a smoothing/interpolation technique), which relies on the raw quantity of data. Who has a copy of the Internet's treasure-trove of text in its data centers?
Ultimately, NLP is something that's Google's to win if it truly desired. I'm not sure what powers Siri directly, but large data is Google's answer, and it's much more effective than Mr. Schmidt would make you believe.
[1] Large Language Models in Machine Translation: http://acl.ldc.upenn.edu/D/D07/D07-1090.pdf