One of my colleagues at the SimCenter: National Center for Computational Engineering threw out the question of what PhD topic to choose. My response was as follows (he’s from Iran, thus my reference to the 1979 Revolution):
The ideal of the PhD topic–which has been held in front of me ever since undergraduate days before the Revolution–is that the topic be original and push the field of knowledge forward.
The problem getting there is twofold.
The first is that, with the speed of advance of technology, by the time you get there it isn’t original any more.
The second is that most academics like topics they have the solution for before they start. With most research being funded, the funding agencies like positive results, not “The proposed method is rubbish” even when that result actually is an advance for the field. That tends to curb real originality.
One thing you might consider is a topic which has been passed by in the flow of research. Science’s path isn’t always a straight line, and isn’t always pushed by “scientific” considerations.
Another is, since you don’t want to say your own proposal is rubbish, find someone else’s that is and show it. That will get emotions hot at technical conferences but, as we say in American politics, the only thing worse than bad publicity is no publicity.
These are some ideas, they may sound absurd, but one never knows, today’s absurdity is tomorrow’s reality.