Well you can just start a startup that works on hard problems, instead of one which revolves around your ajax code rendering correctly on various browsers.
You can solve a hard AI problem with a couple of good hacker friends, just as you could do that at Google. True that it'll probably take you more time than building another website around social networking.
I think that it's unrealistic to build a startup around an unsolved fundamentally hard problem. Can anyone think of a counter-example that succeeded? Typically, the problem is solved first, and then the startup is built to implement the solution.
Solving hard problems depends too much on creativity and research. I'm not sure I could do that under the pressure associated with a startup. The Google environment doesn't seem ideal either, come to think of it, but at least, I'm not the one absorbing the huge risk of failure.
Anybots is a startup that solved a hard problem, but it was self-funded, right? Would YC have funded them if it had been proposed by a 25-year-old Trevor Blackwell?
That's exactly what Andy Brice did with PerfectTablePlan.com
Uses some pretty advanced algorithms for planning seating arrangements.
Of course, the real challenge with a consumer product is usability, that intersection b/t your cool product and the folks with the wallets. But, that's where the money is. (Money isn't everything, of course).
