Levels of Approaches to Simulating the Human Brain
Extreme bottom up approach: Simulate every cell in the brain.
-Semi-extreme bottom up approach: Blue Brain Project where eventually every cell in a
neocortical column is simulated.
Jeff Hawkins bottom up approach: Attempts to simulate cortical functions without
duplicating every variation of neuron itself.
Neural networks modified to be easier for computers
There are two approaches to making artificial general intelligence. The first one is to
program a computer. Since Microsoft is spending so much money and time creating operating
systems which are not even pretending to be intelligent or to learn, I am
starting to think that this is not the way to go. As you point out, there are meetings
and decisions and complexities required. What may be needed is a “blank slate”
neural network which can evolve to have more specific areas such as “speech centers”
and so on. Such a neural network would be programmed much like a human is
educated by experience and then more formal education. The robot beginner,
intermediate and advanced sections of my donbot site are dedicated to educating
robots which are built according to this latter technique. Many of the lessons are in
subjects such as handling human illogic and attraction to myth and fiction.
Artificial Intelligence needs a subconscious if it is to simulate the animal mind.
Consciousness might be the wrong goal if 95% of what humans do is driven by the
subsconscious mind.
Narrow But Deep problems that if you can do them, you must be emulating real intelligence.
Meanwhile inexpensive hardware is coming. PS3 for Supercomputer.
Journal:
http://scientific-direct.net/c.asp?695315&7bc8dfd6ee928076&13
Next
Extreme bottom up approach: Simulate every cell in the brain.
-Semi-extreme bottom up approach: Blue Brain Project where eventually every cell in a
neocortical column is simulated.
Jeff Hawkins bottom up approach: Attempts to simulate cortical functions without
duplicating every variation of neuron itself.
Neural networks modified to be easier for computers
There are two approaches to making artificial general intelligence. The first one is to
program a computer. Since Microsoft is spending so much money and time creating operating
systems which are not even pretending to be intelligent or to learn, I am
starting to think that this is not the way to go. As you point out, there are meetings
and decisions and complexities required. What may be needed is a “blank slate”
neural network which can evolve to have more specific areas such as “speech centers”
and so on. Such a neural network would be programmed much like a human is
educated by experience and then more formal education. The robot beginner,
intermediate and advanced sections of my donbot site are dedicated to educating
robots which are built according to this latter technique. Many of the lessons are in
subjects such as handling human illogic and attraction to myth and fiction.
Artificial Intelligence needs a subconscious if it is to simulate the animal mind.
Consciousness might be the wrong goal if 95% of what humans do is driven by the
subsconscious mind.
Narrow But Deep problems that if you can do them, you must be emulating real intelligence.
Meanwhile inexpensive hardware is coming. PS3 for Supercomputer.
Journal:
http://scientific-direct.net/c.asp?695315&7bc8dfd6ee928076&13
Next
 | |  | |  | |  |