The Chinese Room - Assignment #3

This is not my first exposure to the Chinese room concept. I first had it explained to me a few years ago, but not in near this much detail. I mostly agree with the author in regards to the fact that computers can not reason and understand in the same ways that a human does. What understanding can be obtained when you are only capable of fallowing a simple instruction that is placed in front of you?  Does this exclude an AI from having it own form of understanding though? I believe the reason the Chinese room scenario is still relevant today is because it questions the definitions of the human boundaries of understanding and computing just as much as it questions the AI's ability to truly understand.

One topic I with the article address more thoroughly is when and how a human (computer) goes from not being able to understand the spoken language to being able to fluently communicate. Babies come into the world lacking the ability to either understand speech or express their own desires with any degree of certainty. Since they have no previous understanding of speech or the definition of its words they are forced to rely on context clues they can otherwise since.
"Well the one that provides milk is present 89% of the time I hear the noise 'Mama'. That falls within an acceptable margin of error. I can probably link that noise and that person together in some way." - While it is highly unlikely a newborn thinks like this, the process of natural learning has to set to some metric if we hope to someday understand exactly what has been learned( I do not believe standardized tests will work).  Is the information considered 'learned' after it is utilized and then positively reinforced? If so how is a child leaning to add and subtract any different than a computer's understanding?

I believe the secret of learning new information(and modify old) lies in our ability to introspectively analyze what we already know. Continuing with the Chinese language example, all languages are constantly evolving. As stated in the paper, if  the room had a man in it providing output that was only able to fallow the provided rules(no true understanding of the language) native Chines speakers would be unable to differentiate whether the man inside could truly understand them. We will call this man MAN-2012. MAN-2012 has a finite set of rules and can not deviate from them. As the culture slowly changes outside the room, new words will be created, new slang invented, and the meaning of some words will change with the times. MAN-2012 has no rules in place to handle these changes, and the system breaks down the first time such a change is encountered. I do not know if it is possible for an AI to manage information like this in a real time environment. Each new change would need to change an exponentially increasing number of related topics.