My father-in-law just asked me an interesting question:
How much would an inch of water covering an acre weigh?
It's not such a difficult question. It seems like a perfect query for Wolfram Alpha.
WA doesn't get the question at all. In fact, it seems to have classified an acre as an animal. Maybe I should have said hectare? Nope. It does no better.
Interestingly, Google makes solving this problem trivial, thanks to the useful "convert" command. I solved the problem in two steps. (It helps to know that a cubic centimeter of water weighs a gram - in fact, that's the definition of a gram.)
Convert 1 inch x 1 acre to cubic centimeters -> 102 790 153 cubic centimeters
We already have the answer, 102,790,153 grams, but to put it into terms that are more human I Googled:
and got the correct answer, 113.306748 short tons.
No matter that my father-in-law used a calculator, we both got the same answer, proving that Google is, in some cases, better at numeric analysis than Wolfram Alpha.
The real problem with WA is that it's not easy to formulate a query that produces the results you're looking for. Type 'blood alcohol' and you'll get fascinating results (thanks to Chris Heath for finding that, by the way), but not necessarily results you can expect. And, as it turns out, predictability is an important feature of any search engine, or computational knowledge engine for that matter. Before it's useful you need to have some idea of what kind of answers you might get, and, for the moment, Wolfram Alpha's results seem utterly random. I'm rooting for it, but it may be that it's just too smart for The Rest Of Us.
(Thanks to Edward Coffey who points out that the query "1 acre * 1 inch * 1g / cubic centimetre" works on Wolfram Alpha. Quite well, in fact. But I don't think that changes my point.)