Neural networks and computer simulation

[coberst]

Neural networks and computer simulation

 

Computers are used to simulate hurricanes for the purpose of comprehending hurricanes and also for the purpose of predicting their behavior. Computers are often used to simulate various phenomena for similar purposes.

 

In the effort to utilize computers as controllers for robotics and to simulate human cognition we have developed a great deal of empirical evidence concerning human functioning. The software engineers must develop detailed comprehension of bodies in motion and intellects in action in order to write the algorithms required to build robots of all forms whether to simulate human intellect or human motion.

 

AI (Artificial Intelligence) research began shortly after WWII. Alan Turing was one of the important figures who decided that their efforts would not be focused on building machines but in programming computers.

 

Some of the achievements of AI have, in the last few decades, been turned toward neural modeling. This effort has been taken up by NTL (Neural Theory of Language) research group at Berkeley headed by Jerome Feldman and George Lakoff.

 

The brain’s neural network does things and the task the NTL has set for itself in the 1990s was to discover how the brain does what it does and where in the brain these tasks take place. As of 1999 NTL had decided to undertake three major modeling tasks:

1) The Spatial-Relations Learning Task

2) The Verbs of Hand Motion Learning task

3) The Motor Control and Abstract Aspectual Reasoning Task

 

“In each case, it has been shown that neural structures modeling aspects of the perceptual and motor systems can carry out the given task for concepts, and that, so far as anyone cal tell thus far, those perceptual and motor models are required to carry out the task.”

 

That is to say that the sensorimotor system in the human body can perform the functions required to conceptualize and, infer from those conceptions, in a manner required by human cognition. The logical assumption is that these self same sensorimotor neural networks are the networks the body uses to conceptualize during cognition.

 

Quotes from “Philosophy in The Flesh” by Lakoff and Johnson

[coberst]

“That is to say that the sensorimotor system in the human body can perform the functions required to conceptualize and, infer from those conceptions, in a manner required by human cognition. The logical assumption is that these self same sensorimotor neural networks are the networks the body uses to conceptualize during cognition. “

 

The above paragraph that I have copied from the OP contains the really important conclusions upon which the new cognitive science paradigm is based.

 

 

We have in our Western philosophy a traditional theory of faculty psychology wherein our reasoning is a faculty completely separate from the body. “Reason is seen as independent of perception and bodily movement.” It is this capacity of autonomous reason that makes us different in kind from all other animals. I suspect that many fundamental aspects of philosophy and psychology are focused upon declaring, whenever possible, the separateness of our species from all other animals.

 

This tradition of an autonomous reason began long before evolutionary theory and has held strongly since then without consideration, it seems to me, of the theories of Darwin and of biological science. Cognitive science has in the last three decades developed considerable empirical evidence supporting Darwin and not supporting the traditional theories of philosophy and psychology regarding the autonomy of reason. Cognitive science has focused a great deal of empirical science toward discovering the nature of the embodied mind.

 

The three major findings of cognitive science are:

The mind is inherently embodied.

Thought is mostly unconscious.

Abstract concepts are largely metaphorical.

 

“These findings of cognitive science are profoundly disquieting [for traditional thinking] in two respects. First, they tell us that human reason is a form of animal reason, a reason inextricably tied to our bodies and the peculiarities of our brains. Second, these results tell us that our bodies, brains, and interactions with our environment provide the mostly unconscious basis for our everyday metaphysics, that is, our sense of what is real.”

[LaurieAG]

The logical assumption is that these self same sensorimotor neural networks are the networks the body uses to conceptualize during cognition.

Quotes from “Philosophy in The Flesh” by Lakoff and Johnson

 

It's interesting to note what happens when you apply the basic communications model to the cognitive process utilised by knowledge bases and neural networks. The key difference is that a neural network can enlarge its own zone of experience while a knowledge base cannot.

 

When you dig a bit deeper into the basic mechanics and components of the basic communications model it appears (at least to me anyway) that the traditional use of 'noise' as being an interference to the communication process, was wrong because there could be situations where there was no overlap of the communication context and the respective zones of experience of one or both of the parties in the communication.

 

When you use this type of model for committee and corporate structures you also come up with interesting neural structures.

[coberst]

It's interesting to note what happens when you apply the basic communications model to the cognitive process utilised by knowledge bases and neural networks. The key difference is that a neural network can enlarge its own zone of experience while a knowledge base cannot.

 

I guess that the big difference between the neural network and the knowledge base system is that the neural network can learn without negative feedback whereas the knowledge base system cannot.