They aren't using the Web to power their probability based translation engine. They are using professionally translated books and other texts.

Yeah, I've heard the same thing. I hear they are using a corpus of UN-translated documents.

Probably now, but in the future, they may be using the web. There's just too much data out there to ignore I think :)

Plus, internet language is much more dynamic. Look at how long it took "spam" to be used by leaders in the industry, and think about how long before its put into a professionally translated text.

There's no reason to expect the process to be entirely automated. I expect that even if the Web is used as a data source for translation, some minor human intervention to admit/disallow new elements of language is all that will be required to avoid the danger of potential translation bombs.

I don't think there's anything to worry about.

Google translation work is really dam good.I use this lot's.And it realy helps me alot.
You can visit me at:
http://Spanishwebs.info