It is often said that neuroscience is the study of the mind.
While there is no denying that neuroscientists do study the brain and how it
works, I say that Artificial Intelligence is just as much as about studying how
humans work as it is knowing how the human brain works. Brian Christian spends
a good part of his book talking about how the human mind works and how he
studied this in preparation for the Loebner prize. He wanted to know how to
behave as human as possible (and what that even meant) in the same way that the
programmers who competed wanted to know the same thing to model it in their
submission. This could arguable be more important to the Loebner prize that an
understanding of programming. While there would be no program without the
programming knowledge, the computer would have no chance of winning if the
programmer did not have an understanding of human behavior and how to model it.
One of the main ways to make advances in Artificial Intelligence is to have a
better understanding (and therefore, become better able model) of human
behavior and human tendencies.
Compare the Loebner prize with the Deep Blue chess
competitions. Deep Blue was programmed to play chess by the programmers
studying the theory behind how chess plays out and the patterns their opponent
had used in past matches. The programmers had to work to model the logic that
goes into chess matches and all of the theory that the masters use when they
are preparing for matches. The programmers for the Loebner do the same thing,
except they study a more complicated set of the logic, the logic that dictates
the basics of human behavior and conversation. By trying to replicate these
patterns and logics, we are essentially trying to nail down the why behind what
we do. Successfully creating a program that models human behavior will not only
create another “species” of equal intelligence, but it will actually be some
sort of mirror into our own nature. This prospect may be even more frightening
then the idea of creating what may one day destroy us, but that it will be our
own concept of us. Our own reflection will bring about our downfall.
I think the concept you mention of "trying to nail down the why behind what we do" is an interesting one. Mainly because it makes me wonder: if we ARE able to succeed at "nailing down" the definition of a human, what are the implications of that? I suppose it seems that this would indicate something negative about humanity, but what exactly I am not sure. Obviously I think we want to believe that we cannot be "nailed down" and that that is why we say we HOPE that computers cannot ever exactly resemble us. But if they COULD one day do so, would that be the end? I mean...it would still be the case that we CREATED them, so just by virtue of being their intellectual predecessors, couldn't our uniqueness or sense of being special still be somewhat maintained?
ReplyDeleteThis post sounded a lot like Dostoevsky to me, except that he wouldn't consider the possibility of a program modeling human behavior. For the program to ever exist, and we actually could nail down the definition of the human, then it would appear we wouldn't be free.
ReplyDeleteAt first, Karissa, I wanted to agree with your point that we would still have some leg up on computers if they ever completely replicated our consciousnesses in virtue of us being their creators, however I don't think this is much of a substantial difference, because if we ever successfully created an artificial intelligence that replicated ourselves, then they would have the capacity to likewise be creators of further intelligence. Our identity in this hypothetical future as "the creators" would quickly be eroded when A.I. begins to create, and a mixed society with this notion of "original creators" as a hierarchical difference would be dismantled just as parents are no more intrinsically legally valuable than their offspring.
ReplyDeleteI think it's important for people who work on AI to know about how humans work. However, this might not be fully to make computers function like humans. It might be to make computers function like humans in some good, useful ways while doing away with the other functions of being human. The most successful AI might be a computer that keeps all the positive human functions while also disregarding the functions of the human brain that aren't useful or that are counterproductive. I think this is why some people at least view computers as one day being better than humans at being human.
ReplyDeleteI find it incredibly interesting that you claim "Successfully creating a program that models human behavior will not only create another “species” of equal intelligence, but it will actually be some sort of mirror into our own nature. This prospect may be even more frightening then the idea of creating what may one day destroy us, but that it will be our own concept of us. Our own reflection will bring about our downfall." Forgive me if I'm misunderstanding you, but it seems like you are articulating a point that Heidegger makes in an essay called "The Question Concerning Technology". In this essay he argues that technology has changed our way of thinking and that technological thinking essentially tries to dominate any other modes of thinking. It is precisely this kind of technological conception/reduction of ourselves that threatens us. Because of space and clarity I'm going to put an analysis of Heidegger's argument aside and focus on the idea that several of us have mentioned: that computers will be superior. First off, I'd like to point out that while computers are definitely more efficient at crunching data and the like they are better because of the work and discovery of humans. How do we characterize our relationship to our inventions? (I'd like to limit the scope to current technology since we have no rational way to determine what the scope of technological advancement will be) It seems rather obvious that there is a difference between a computer-scientists relation to her program and her son but what is the quality/characteristic that makes the difference?
ReplyDelete