For those of us who have learned a new skill we know that when we learn the wrong technique in let us say sports and once that technique is committed to memory, it is extremely hard to un learn. However, this may not be the case with training artificial intelligence because all we need is a simple program, which reverses or erases the imprinted data sets, which are incorrect. The human imprinted memory can be changed by learning a new way of doing something, which is ironic to the brain, as it is different from what was committed to memory, thus allowing a memory overwrite. In artificial intelligence the imprinted data sets will be coming in from sensors and during evaluation of the decisions rendered by a decision matrix that has imprinted the data. If the decision which was made based on this data did not yield the proper results expected or desired, then the artificial intelligence program must initiate a subprogram to reverse itself through the data used, working backwards to find the flaw or which decision was in error and then erasing back to that point. In fact it may be easier and simpler to teach a computer to learn, then add an actual organic brain. And as any athlete knows committing one's motions and techniques to muscle memory can take out to 10,000 or more motions before it becomes a reflex. But in artificial intelligence, we may be able to fix the problem in one quick erasing event and starting over. Too abstract for you, perhaps, but it is possible and therefore should be discussed, so consider this in 2006. |
