Analyzing the activity of large populations of neurons: how tractable is the problem?