Neuron network design


From the outset, both computer technology and the view of the brain it inspired have been based on the classic von Neumann architecture, which involves commands executed serially, or step by step. In contrast, new "connectionist" models in computer science---especially in the field of artificial intelligence---are inspired by recent developments in neuroscience, the study of the brain's neurons.

It turns out that the biochemistry of neurons is neither digital nor serial. That is, each neuron has many levels of connection with other neurons, some more "excited" and others more "inhibited," rather than the simple on/off connection that characterizes a digital link. And the connections it makes go in many directions at once, creating a local network that can vary with the intensity of the energy transmitted from cell to cell.

As might be expected, the more pronounced the pulse the larger the network that is created. A strong pulse may create a wider network, involving more neurons in creating the "memory"; a strong inhibitory signal may lower the level of interaction over a large number of neurons. Memories are not stored at a particular site, apparently, but are created via shifting fields of interconnections.

Connectionist computer modelling, similarly, designs large arrays of simple units, analogous to neurons. Each has numerous excitatory or inhibitory connections to other units, as neurons do. The levels of activity such connections display can vary more greatly than the standard digital link, and many connections can be made at once via the other links in the local network. The processing in such new computer systems is parallel rather than serial, and each computational decision (the equivalent of whether a synapse will create a link with another neuron or not) is based on "local" connections within specific networks; it is not centrally directed or autonomously self-contained.

Where all this will go, no one knows.

[1989]

Batch

Processing




Compare:

"[E]very age has its mythical figures that transgress the boundaries it creates between the human and the non-human, culture and nature."

quoted in The Cyborg Handbook, Chris Hables Gray, ed., with the assistance of Heidi J. Figueroa-Sarriera and Steven Mentor. NY: Routledge, 1995.


back to Node 3 contents list