Oleh: microbatoe | Juli 11, 2008

Artificial Intelligence

Artificial intelligence
People dream of creating a machine with artificial intelligence (AI)
that rivals or surpasses human intelligence. I feel neural networks
are the best technology for developing and generating AI in computer
systems. This is in contrast to other computerists who see
expert systems and task-specific rule-based systems (programs) as
potentially more viable.
It is an undeniable fact that rule-based computer operating systems
(DOS, Windows, Linux, etc.) and rule-based software are valuable
and do most (close to all) of the computer labor today. Even so, the
pattern matching and learning capabilities of neural networks are
the most promising approach to realizing the AI dream.
Recently it had been forecasted that large-scale parallel processors
using a combination of neural networks and fuzzy logic
could simulate the human brain within 10 years. While this
forecast may be optimistic, progress is being made toward
achieving that goal. Second-generation neural chips are on the
market. Recently two companies (Intel Corp., Santa Clara, CA,
and Nestor Inc., Providence, RI), through joint effort, created a
new neural chip called the Ni1000. The Ni1000 chip, released in
1993, contains 1024 artificial neurons. This integrated circuithas 3 million transistors and performs 20 billion integer operations
per second.
Evolution of consciousness in artificial intelligence
Consciousness is a manifestation of the brain’s internal processes.
The generation of consciousness in Homo sapiens coincides
with the evolution and development of neural structures (the
brain) in the biological system. A billion years ago the highest
form of life on Earth was a worm. Let’s consider the ancestral worm
for a moment. Does its rudimentary (neural structure) intelligence
create a form of rudimentary consciousness? If so, then it’s akin to
an intelligence and consciousness that can be created by artificial
neural networks running in today’s supercomputers (see Fig. 2.1).
In reality, while the processing power of supercomputers approaches
that of a worm, this has not yet been accomplished. The reason is that
it is too difficult to program a neural network in a supercomputer
that would use all the computer’s processing power.
The worm is unquestionably alive, but is it self-aware? Is it simply a
cohesive jumble of neurons replaying an ancestral record imprinted
within its primordial neural structure, making it no more than a
functional biological automaton?
Chapter two

Team LRN