Wolfram Alpha is indeed a search engine, but it's not pretending to be Google, and unlike so many of the search engines that I've had pitched to me over the years, it isn't trying to do "search". It's actually doing something more subtle: it's doing semantic search. That thing that Sir Tim Berners-Lee has been banging on about.
Whereas Google simply uses tweaked versions of its original "lots of people link to this page" algorithm.
So don't write it off just yet. In fact don't write it off at all. The only thing that might hold it back is the cost of running it - but as Wolfram has had huge success with its software package Mathematica (which it used to build Alpha), it may be able to sustain the cost for a while.
The potential benefits mean that Wolfram Alpha could become both more reliable than Wikipedia for straight factual questions (though it will never be as in-depth as Wikipedia), and that its usefulness will grow very rapidly as more and more pages on the web get the sort of XML markup that means they can distinguish between Ford, the car, and ford, the method of getting across a river.
It's as though the enormous overpromise made all those years ago by Ask Jeeves - that it would understand natural-language queries - is finally starting to come true.
I think Wolfram Alpha has been incorrectly positioned and messaged. From the early reviews (Tech Crunch, RWW, Mashable), I heard it would be a "Google killer," so I asked it a few Google type search questions to no avail.
Now I'm trying to figure out what to ask it that's appropriate. It's like trying to feed a picky eater.
Posted by: Francine Hardaway | May 21, 2009 at 06:44 PM