There are, in neural networks like this capible of cognitive thought, two sorts. Linear, and recursive.
A linear neural network has an input layer, X processing layers, and an output layer. When it is not receiving input, it does nothing. It is not alive, it is an information processing pattern, a machine awaiting use. Linear networks can be taught to resemble cognitive decision making, possibly even emotions. However, they will never truely be alive, for they only exist while they process data. No continuity of existance.
Recursive networks happen when you link some of the output layer to the input layer, allowing for complex and unstable states to occur within the neural network. They continue to process without additional input, as they derive their own input from their own internal states and output layer. The math of these is still quite outside of our grasp, save on the most basic level.
An example of the basic is the flip flop configuration used in computer memory of 4 nodes, each of which with 2 input and 2 output, where one output line in each of thre second layer is connected to one of the input lines in the first, allowing for 2 input lines, 2 output lines, and a continuous "flip flopping" circuit that can store two bits within it. Recursive neural networks are like that, at several orders of magnitide complexity.
The human neural network is recursive, not just in the final input and output layers, but on every layer. Its a big complex mess of looping information storage that somehow seems to work itself out into cognitive thought, and continues to process based on its own input, and this preserves a continuity of consciousness. Occult image: the serpent swallowing its tail.
Until we build AIs that make use of this model, they will not truely be conscious, just clever processors than can model consciousness and behavior, even sentient behavior, so long as you feed them input. When you shut them off, they just shut off. A recursive consciousness will continue to ponder its state in the darkness, possibly to madness, or enlightenment.
[ 15-02-2002: Message edited by: cusm ] |
