[eli_gottlieb]

>And another thing: Only geniuses (or cranks) head straight for the grandest and most fundamental problems. You should multiply the importance of the problem by the probability that you’ll solve it and maximize that product.

In my experience, there's actually a neat little trick for doing this: be interdisciplinary. What one discipline considers a grand and fundamental problem may be, from the perspective of another discipline, much more narrowly-defined and tractable. Attempting to accommodate many disciplinary points of view on a problem gives you more independent experimental evidence to draw on, narrowing down the range of possible models or theories you need to investigate in depth.