Microsoft Word Kurzweil, Ray The Singularity Is Near doc

Modeling Regions of the Brain 
Most probably the human brain is, in the main, composed of large numbers of relatively small distributed 
systems, arranged by embryology into a complex society that is controlled in part (but only in part) by serial, 
symbolic systems that are added later. But the subsymbolic systems that do most of the work from underneath 
must, by their very character, block all the other parts of the brain from knowing much about how they work. 
And this, itself, could help explain how people do so many things yet have such incomplete ideas on how 
those things are actually done. 
—M
ARVIN 
M
INSKY AND 
S
EYMOUR 
P
APERT
73 
Common sense is not a simple thing. Instead, it is an immense society of hard-earned practical ideas—of 
multitudes of life-learned rules and exceptions, dispositions and tendencies, balances and checks. 
—M
ARVIN 
M
INSKY
In addition to new insights into the plasticity of organization of each brain region, researchers are rapidly creating 
detailed models of particular regions of the brain. These neuromorphic models and simulations lag only slightly behind 
the availability of the information on which they are based. The rapid success of turning the detailed data from studies 
of neurons and the interconnection data from neural scanning into effective models and working simulations belies 
often-stated skepticism about our inherent capability of understanding our own brains. 

---

Modeling human-brain functionality on a nonlinearity-by-nonlinearity and synapse-by-synapse basis is generally 
not necessary. Simulations of regions that store memories and skills in individual neurons and connections (for 
example, the cerebellum) do make use of detailed cellular models. Even for these regions, however, simulations 
require far less computation than is implied by all of the neural components. This is true of the cerebellum simulation 
described below. 
Although there is a great deal of detailed complexity and nonlinearity in the subneural parts of each neuron, as 
well as a chaotic, semirandom wiring pattern underlying the trillions of connections in the brain, significant progress 
has been made over the past twenty years in the mathematics of modeling such adaptive nonlinear systems. Preserving 
the exact shape of every dendrite and the precise "squiggle" of every interneuronal connection is generally not 
necessary. We can understand the principles of operation of extensive regions of the brain by examining their 
dynamics at the appropriate level of analysis. 
We have already had significant success in creating models and simulations of extensive brain regions. Applying 
tests to these simulations and comparing the data to that obtained from psychophysical experiments on actual human 
brains have produced impressive results. Given the relative crudeness of our scanning and sensing tools to date, the 
success in modeling, as illustrated by the following works in progress, demonstrates the ability to extract the right 
insights from the mass of data being gathered. The following are only a few examples of successful models of brain 
regions, all works in progress.