I have spent a lot of my time contemplating "big ideas" both when I was a PhD candidate (various set theory proofs, signal processing ideas) and in my own software (machine learning methods and application to gambling etc) and I find that mostly thinking about this stuff is just mental masturbation. I typically found some flaw later on and I would have been much better off with more manageable problems that gave more direct, smaller, higher probability of success profits. Basically, I think if you don't have the resources to implement it, it's not a good idea for you. Of course, if you have some amazing insight into a hard problem (how many $$$$$ problems are out there that haven't had serious attention from seriously smart people?) then you can build a prototype and seek funding but I would be very very cautious.