Without seeing a demo, it's hard to say anything factual - but in principle, the "problem" in any well-design/near-perfect system is still human beings. (context, language, HCI, "true", "news" etc) So while it is possible that Wolfram Alpha has successfully modeled a completely new area of search/information modeling - more likely, WA is another novel way parse language and information.

Also, Freebase and Wikipedia might the competitors (I agree with Nova Spivak) - but Wolphram Alpha won't be a panacea until it's open source data/technology. The wisdom of the crowds is the only way to model... the wisdom of crowds.