Microsoft Word Kurzweil, Ray The Singularity Is Near doc

Neuron Models 

---

Researchers are also discovering that specific neurons perform special recognition tasks. An experiment with chickens 
identified brain-stem neurons that detect particular delays as sounds arrive at the two ears.
60
Different neurons respond 
to different amounts of delay. Although there are many complex irregularities in how these neurons (and the networks 
they rely on) work, what they are actually accomplishing is easy to describe and would be simple to replicate. 
According to University of California at San Diego neuroscientist Scott Makeig, "Recent neurobiological results 
suggest an important role of precisely synchronized neural inputs in learning and memory."
61
Electronic Neurons.
A recent experiment at the University of California at San Diego's Institute for Nonlinear 
Science demonstrates the potential for electronic neurons to precisely emulate biological ones. Neurons (biological or 
otherwise) are a prime example of what is often called chaotic computing. Each neuron acts in an essentially 
unpredictable fashion. When an entire network of neurons receives input (from the outside world or from other 
networks of neurons), the signaling among them appears at first to be frenzied and random. Over time, typically a 
fraction of a second or so, the chaotic interplay of the neurons dies down and a stable pattern of firing emerges. This 
pattern represents the "decision" of the neural network. If the neural network is performing a pattern-recognition task 
(and such tasks constitute the bulk of the activity in the human brain), the emergent pattern represents the appropriate 
recognition. . 
So the question addressed by the San Diego researchers was: could electronic neurons engage in this chaotic 
dance alongside biological ones? They connected artificial neurons with real neurons from spiny lobsters in a single 
network, and their hybrid biological-nonbiological network performed in the same way (that is, chaotic interplay 
followed by a stable emergent pattern) and with the same type of results as an all-biological net of neurons. 
Essentially, the biological neurons accepted their electronic peers. This indicates that the chaotic mathematical model 
of these neurons was reasonably accurate.