Yet another brilliant analysis by Cory. And yes, as a computer vision researcher, I've been annoyed at the 'throw more data and power at the problem' type of data. Yes, it works, but constraining the problem and working within those constraints is a lot more interesting.
