PERSON
David Marr
The neuroscientist who died at thirty-five assembling a book about vision and left behind the most rigorous framework ever devised for asking what any information-processing system—brain or machine—is actually doing.
David Marr completed
Vision in the shadow of the leukemia that killed him in 1980, and the book arrived two years after his death like a letter from the future. Its subject was how the brain turns light into sight; its lasting contribution was a method for understanding any system that processes information at all. Marr argued that such a system must be understood at three distinct levels: the
computational, the algorithmic, and the implementational—what problem is being solved and why, by what procedure, and in what physical stuff—and that these levels are genuinely separate, so that answers at one cannot substitute for answers at another. His central, unfashionable claim was that the highest of them, the question of what problem is being solved, is the one most often skipped and the one that matters most. Trying to understand perception by studying only neurons, he wrote, is like trying to understand bird flight by studying only feathers: it just cannot be done. That diagnosis describes