Plus, I think today's searchers - for the most part - speak search engine lingo, and have certain expectations. I don't want my search results to play D# major if that's what I'm searching for, I probably want to see the notes. I won't type in "10 flips for four heads", I'll type in "coin tossing probability".
I can see kids doing homework projects having fun with this, but beyond? Dunno. I'm quite happy for it to be a wild success and prove me wrong, I'm just very sceptical this will happen.
That is the tough issue here. The hype has been immense.
In other words: Google can help you find things like a phone book can help you find things, Alpha can help you find things like a calculator can help you find things.
I'm sure Wolfram could partner with Amazon to crowdsource some of this data gathering through Mechanical Turk. For example, collecting data on celebrities (e.g. height, birth date, family relations). Turk users would fill out forms and cite where they got the data from.
Additional note: I get super pumped when I read something on HN and I think to myself "Damn! That is a fine idea." It always makes me grin.
The ontology of Wolfram Alpha is involved. Those professional researchers in Champaign who understand how to talk to Wolfram Alpha are not getting their data from simple Google searches and then citing their sources. This is a complex process even for simple data points. If Wolfram Alpha has a need for Mechanical Turk it would be to catch mistakes or maybe do information triage - but Stephen made clear he doesn't want quasi experts going anywhere near the research department.
http://www.wolframalpha.com/ still says "Launching May 2009" and it's now May 3rd.
Seriously, though - I can't help but think that no matter how brilliant Wolfram Alpha might be, it will never end up being as good as he thinks it is. I hope I'm proved wrong, but that unpleasant feeling has definitely coloured my opinion thus far.
