Again, "human input." That would be the end users making their queries in this case. They would seem to be the equivalent of photographers operating their cameras here, so I'd expect that the end users own the copyright on the output.

Reading http://laws.findlaw.com/us/499/340.html, it's clear that there's a "creativity" requirement for copyright, and that only tangible instantiations of a work are copyrightable -- not general principles or algorithms. Wolfram's system is on full autopilot and would seem to be the analog of a camera in this discussion, and my understanding of the current state of technology and law is that only humans are capable of creativity.

Wolfram's software is surely copyrightable, as are certain human-created elements (e.g. their logo) which are copied into the final output. The human user's formulated query is probably copyrightable "authorship." However, Wolfram's system has no humans in the loop and a mere mechanical process cannot change authorship, so it would seem that the user and Wolfram both own copyrights on different parts of the output. The user probably owns the "meat" of the results, much as if I would own the copyright if I rented Wolfram's camera to take photos.

Anybody can claim copyright on anything, but making it stick is another matter.

Machine-generated output per se isn't copyrightable, since machines aren't (yet) capable of original authorship. Of course, computer output is copyrightable if it also contains original, human-generated content, for example Wolfram's logo, etc. See http://en.wikipedia.org/wiki/Copyright, search for "authorship."

I have seen cases where patents are issued for "inventions" that were allegedly previously published in big-name publications (e.g. IEEE Transactions on Information Theory).

As a real-world matter, getting and abandoning a patent has a significant advantage over other methods of publication: if somebody else tries to patent your invention later, the Patent Office has a better chance of rejecting their application. The fact is that the USPTO examiners are pressed for time and usually only search their own patent and patent application databases for prior art because it's quick & easy for them. Anyone who's ever been involved in a patent application will remember that the examiner raises prior-art objections by throwing patents or apps back at you, not journal articles or websites.

Obviously, helping the USPTO to reject a patent in the first place is much better than the expense, time and anxiety of trying to overturn somebody else's patent later. Naturally there's no guarantee that the examiner's search will actually turn up your patent or application, but being in their data base improves your odds significantly.

Don't bother to RTFA, everybody here already knows everything in that article.

There's a much safer way of charging your EV from overhead power cables for free. People have been doing it to heat their homes (illegally, of course) in outlying areas of Canada for years: place an inductor under a 350kV powerline & run wires to your house. The powerline operators hate it because the inductance messes up their power factor, and the poachers eventually get caught because the powerline operators sporadically use a small plane to patrol their rights of way for inductors. But, a 10 minute recharge-and-flee time would make detection near-impossible!