Patterns and Randomness

In the Hayle's reading "Virtual Bodies and Flickering Signifiers" she writes (on page 78) that, "the interpaly between pattern and randomness causes the system to evolve in a new direction." I assume that when she uses the word system she implies both the human body and technological systems as she has implied earlier in this essay. Doesn't a computer run completely on algorithms? And a computer cannot acct on codes or commands that it wasn't trained to respond to. To add new material to a computer, one must reprogram it with a new set of directions. So a computer (or any other piece of technology for that matter) cannot evolve because it can't "grow" from the mutation. What then does she mean when she says that randomness causes "a system to evolve?" I don't really see how randomness is a part of the modes of new media.

i'm not sure that this is what you mean by a computer "evolving," but computers can certainly be said to "learn." people can train neural networks to do a lot of crazy things (they're called "neural networks" because they mimic the way our own neural networks function). face recognition, i'm pretty sure, is done entirely by neural nets. but let's take a much simpler example for starters.

basically, there is an "input" to the net, which consists of a certain number of numbers, and an "output." along the way, the net multiplies the numbers by some amount (between 0 and 1 i'm pretty sure). if the output is "1" then the input is "approved" and if it's "0" then the input is not approved.

let's say that we want a net that will identify only inputs that consist of the same number, like "1" "1" and "1" but not "1" "2" and "3." We set up the factors (whatever the net will multiply the numbers by) pretty randomly, and then give it training data to practice on. so we give it a series of inputs and also what we want the correct output to be for those inputs. the net will then alter its factors so that it arrives at the correct solutions, and in this way "learns" to recognize certain types of inputs.

wow, that was probably really confusing. anyway the point is that a neural net can start out not being able to recognize faces and then "learn" how to do it. if you want a better, more detailed explanation, you can read about it on wikipedia.