http://www.kurzweilai.net/neural-networks-promise-sharpest-ever-images

Imagine applying this to everyday audio/video via something like Google Glass. Imagine refining huge samples such as a novel or a physics compendium.

Maybe we already have enough data, given this methodology, to describe and extract a full blown universal TOE.  If so, why do anything more?  In general, if we know that this is theoretically possible then what incentive remains, of any kind?  Do we have to...