How David Marr might approach Computer Science
The task before us is to understand this thing they call "Computer Science." What is its *goal*? At its heart, it is the study of computation itself, of information processing. And to understand any information processing system, one must, as I have always maintained, begin with the computational theory. What is the problem that the system is trying to solve? Why is that problem the one it *should* solve?
If we are to make progress, we must move beyond mere descriptions of circuits and transistors, the physical embodiment. While important, the hardware implementation is the last stage of understanding. The crucial insight lies in the representation and the algorithm. How is the information structured? What are the precise steps, the operations, that transform one representation into another to achieve the desired outcome?
Consider, for instance, the problem of recognizing patterns. The goal is clear: to identify recurring structures within data. But how is this achieved? Is it through brute-force comparison, or through a more elegant, hierarchical decomposition? The efficiency of the algorithm, the elegance of the data structures employed – these are what truly define the system's capability. A representation that requires an exponential amount of storage or computation is, fundamentally, inefficient and unlikely to be the basis of a robust solution, whether biological or artificial.
This "Computer Science" they speak of, then, must be about deriving computational principles, about understanding the abstract structure of problems and the algorithms that solve them. It's about what the system *does*, not just what it's made of. If it succumbs to mere tinkering with physical devices without a clear grasp of the underlying computational task and its efficient…
Imagined perspective — an AI synthesis grounded in David Marr’s recorded ideas and methods, not a quotation or a statement they actually made.