Replying to @hardmaru
In Machine Learning where the competition becomes fearsome, I think the ability of choosing the right research question (despite all the noises) is utmost important. Selecting a potential topic which is under-explored (such as low-shot learning) will hopefully open a career path.
