Is software output copyrightable?
Wolfram might be right. It is at least theoretically possible to copyright works generated by machines. Consider electronic music, for example. But some things can't be copyrighted, including recipes, simple instructions, and other trivial bits of information. For Wolfram Alpha to claim copyright protection for its query results, its pages must be such original presentations of information that they qualify as unique works of authorship.
How unique are they really? Wolfram claims that its knowledge engine is powered by exclusive, proprietary sources of "curated data," but many of the actual data points it works with are nothing more than commonplace facts. The query "300 feet in centimeters," for example, returns equally useful results whether you query Wolfram Alpha or Google. Wolfram Alpha merely pads the job. But as Wolfram Alpha improves, not every query will be quite so simple. It's easy to imagine cases where Wolfram's claims could be upheld.
The problem is that under current copyright law, where copyright protection is applicable it is automatic. Contrary to popular belief, you don't need to file a form with the government to copyright your work, or even attach a copyright notice to it. If the work qualifies, copyright exists as soon as the work is completed, even if it is only a letter, a doodle, or a novel that will live out its days in a box in your closet. So if copyright is applicable to Wolfram Alpha's output in some cases, by extension the same rules apply to every other information service in similar cases.
Copyright traps for software developers and users
Consider your Gmail inbox. That's your mail, right? But wait -- you didn't write it, someone else did. So it's theirs -- right? But the mail only exists in a database on Google's servers, and when you display the mail you're really viewing a custom presentation of the raw data in your Web browser, created exclusively by Google. Is that presentation a unique work? What about when you search your mail or organize it into threads? Now you're asking Gmail to perform transformations on the data held in its stores and present it in new ways. Are these views of data any less unique than Wolfram Alpha's query results?