Neural Nets and Artificial Neurons:
Neural network limitations:
Which neural network
implementation do we use?
How do we implement a
network of networks?
Neural networks can't
presently encode relationships.
How do we fabricate a
500,000,000,000,000-synapse net?
The human brain and the
human mind are enormously complex.
One of the questions that might be asked is: why not use neural nets to create artificial intelligence? My reasons for not embracing neural networks at this time is fourfold:
(1) It seems to me that we
are a long way from fabricating 100,000,000,000 neuron nets with
thousands of interconnections per neuronperhaps 500
trillion synapses. The interconnections would be a plumber's
nightmare. Also, the brain possesses the ability to grow new
dendrites and establish new synapses. Of course, artificial
neurons may be able to operate orders of magnitude faster than
biological neurons.
A small New
Mexico State University spin-off called Intelligent Reasoning
Systems, Inc., in Austin Texas, has developed what the company
claims is the closest analog to an artificial neuron yet devised.
It uses a hybrid analog/discrete representation that is said to
more readily tackle such time-dependent tasks as speech
recognition or motion detection without the complications of
digital encoding. (The sampling of speech at regular intervals
fails to encode the natural rhythms of speech, which continually
varies its cadence.) The use of hybrid temporal processing
elements (HTPE's) that are driven asynchronously by their inputs
and that encode information temporally has shown that speech
segmentation occurs naturally at a certain stage in the chain of
signal processing. The developers, Mark Deyong and Randall
Findley, state that silicon-based circuitry is much more reliable
than biological neurons, which have are error-prone and have a
high failure rate. In natural circuitry, there is no guarantee
that a given neuron will fire, and nature makes up for that with
a high degree of redundancyunnecessary for silicon-based
circuits. In addition the 1,000,000 to 1 speed advantage which
silicon circuits enjoy over biological circuits can be used to
further reduce the neuron count. The HTPE approach, because it
more faithfully simulates the behavior of actual neurons, is said
to be well suited to handling time varying, as well as
spatially-varying patterns, in contrast to conventional neural
networks which utilize static (though trainable) weighting
functions. The behavior of, and programming for biological-neuron
analogs (HTPE's) is fundamentally different, leading to networks
which are very sparse compared to those of conventional neural
networks. The HTPE approach requires 7 FET's (Field Effect
Transistors) per neuron and 5 FET's per synapse.
The problem
that I foresee in trying to build an analog of the human brain
using HTPE's or any other collection of neural networks in the
near future lies in the number of synaptic weighting factors that
would have to be represented. Even if HTPE's can remove the
redundancies of natural neural networks and the 1,000,000-to-1
speed advantage of silicon circuitry could lower the neuraon
count to 10,000 or 100,000, the synaptic transistor count (of the
order of 100 trillion?) would seem to be be far beyond
present-day technology. Of course, in the future, given chip
counts with trillions of transistor-like processing elements,
such networks might be feasible. It is even possible that
supercomputer-class systems containing 100 trillion transistors
might be constructed in the year 2000-2005 time frame. (The
planned Cray T3E would house up to 4 trillion transistors.) Such
an investment might be well worth its cost as a
proof-of-principle demonstrator.
(2) The brain apparently contains a multitude of very complex and highly specialized areas which we probably haven't yet fully mapped out, and don't understand. I'm thinking that the functions that I'm discovering probably need to be performed no matter how one goes about it. Experience with biological systems in general shows that they are exceedingly complicated, with multiple backup systems. My consideration of the functions of the mind suggests that it is also extremely complicated. Perhaps this attempt to identify mental functions might even shed light on the specialized areas of the brain. Then, too, it isn't necessary that we exactly emulate the human brain. There may be other ways to achieve similar goals using conventional computers. For one thing, our robot need not be concerned about survival in the wild, and evasion of predators. If we can achieve even a tiny fraction of what the brain can accomplish, we might still have produced a very commercially-useful product.
(3) I don't have access to cutting-edge neural network research facilities. On the other hand, PCs are ubiquitous and I'm thinking that it may make sense to see what we can do with them using conventional, von Neumann programming.
(4) It should be easy to
duplicate the (humongous) file that represents a robot's data
base, using conventional disk-stored files. This should make it
possible to clone the robot, and should also confer
near-immortality upon it. However, reading out and duplicating
the synaptic settings of a neural network might be more
difficult.(?)