Monday, February 25, 2013

Gary Marcus: The New Yorker "Obama's Brain" Excerpt

Magnified 400 times, this is a 2-Photon fluorescence image of glial cells in the cerebellum. Glial cells provide support for the brain's neurons. This image was made by Thomas Deerinck of the National Center for Microscopy and Imaging Research, University of California, San Diego.
At some level, the brain is a kind of computer. It takes in information, combines new information with previously acquired information, and performs actions based on the results of those computations.

Yet we know remarkably little about how the brain performs its computations and how those computations relate to behavior, especially in comparison to what we know about computers.

Computers are, for all practical purposes, entirely understood. We know what they’re made of, we know how electrons move inside of them, and we know how their basic logical functions (such as “and,” “or,” and “not”) combine to form more complex operations, such as arithmetic and control (e.g., performing process X if a password is entered correctly and process Y if the password is entered incorrectly).

 In turn, programming languages relate machine language to more abstract sets of instructions that are more readily understood by human beings. An unbroken chain of inference connects the plans of the programmer to the actions of the electrons that ultimately implement them.

When it comes to the brain, we know comparatively little, in part because the brain is considerably less straightforward. By some counts (though the exact number remains elusive), there are hundreds of different kinds of neurons, each with different physical properties and ways of interacting with other neurons.

We don’t, for example, even know whether the basic unit of computational currency in the brain is digital (e.g., a set of zeroes and ones, like in virtually all modern computers) or analog (like the continuously moving second hand in an old-fashioned clock, an approach that was commonly used in some pre-Second World War computers).

Likewise, although we know something about which brain regions participate in the processing and storage of memories, we still don’t understand how the brain encodes those memories. And we know very little about how the brain’s basic units organize into larger-scale circuits, and how those circuits, often in physically disparate parts of the brain, work together to produce unified behavior.

Modern biology was launched in large part by three discoveries: Oswald Avery’s discovery of DNA; Watson and Crick’s deciphering of the physical structure of the DNA; and the discovery, by several researchers in the early nineteen sixties, of the code by which different triples of RNA nucleotides are turned into amino acids. All three of these discoveries were made possible in part by Mendel’s experiments with peas, in which he identified what we now know as genes.

The brain will finally give up its secrets when we do something similar, by constructing an inventory of the brain’s basic elements—the neural analogs to transistors, logic gates, and microprocessors — and relating those basic elements to broader-scale cognitive computations.

To connect brain to behavior, we don’t need to build a whole brain, as the E.U. aims to do; we need to understand how the brain’s parts work together.

And new techniques, like optogenetics (which allow experimenters to control brain activity—and hence an animal’s behavior—by exposing neurons to light) and fluorescent imaging (which makes it possible to monitor the responses of thousands of neurons simultaneously in awake, behaving animals), make addressing such questions potentially feasible for the first time.

Read more:

No comments:

Post a Comment