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.