....Google has to build a personalized classifier for each Gmail user and it needs a lot of messages. 'Email importance ranking works best for people who receive a lot of email,' explains Google. Google takes into account implicit signals like: the messages from people you frequently email are important, if a message includes words frequently used in other messages you usually read then it's probably important, the messages you star are probably more important than the messages you archive without opening. There are also explicit signals: click on the important/unimportant buttons, create filters to mark messages as important."...Classifiers are used in speech recognition, Google Search and so on -- they're not new. For that matter, this kind of "what's important" ranking has been done many times without using formal classifiers.
Still, this is a mini-milestone of sorts. Classic AI technology is moving to our personal lives from multiple directions. The next step is deploying these classifiers into Google's Facebook-alternative, and then it's another step to our pending lifestream classifiers (which might help keep memory-impaired boomers functional longer).
Yes, those are Skynet's footsteps you hear ... :-).
Yes, this is part of why the IT driver of our whitewater world is not going away, but only getting bigger.