The ultimate goal of most, if not all, classes at MIT is to eventually be able to take the skills and concepts learned to the real world and apply them there. So why is there a fundamental disconnect between what we learn in classes and what is useful in the real world? A perfect example is the structure of many tests. As of now, it is hard to imagine a future where one would not have access to Google, Wikipedia, and other quick and easy sources for equations or worked, simple examples. A test that asks you to memorize these equations is not preparing you for the real world. Instead, tests should focus on applying these ideas to new problems. Obviously the tests must be solvable and have reasonable answers, but the bias of writing "simple" tests that simply require you to state information needs to be done away with. Instead, tests should focus on asking conceptual questions - tougher to grade, but much more relevant to being able to apply those concepts to the real world, where you are likely to become far more specialized in terms of math, and the key will be a broad understanding of the underlying concepts more than a knowledge of equations.
(Another great addition would be more labs and experimental demos, although quality of these is vastly more important than quantity.)