(This page last updated 25 November, 2004.)
|
|
Supplemental ReadingA classic article by Alan Turing is reprinted as Chapter 4 in the book The Mind's I. Turing wrote this article in 1950, when computers were still very new, and most people had very little experience with them. He posed the question "Can machines think?", and although he did not have a definite answer, he proposed a method for answering it, which is now known as the Turing Test. Right now no computer program can pass the Turing Test very effectively, although some can come close in a short version of the test. If a computer program did pass the test, would you conclude that it was capable of thought? That it was intelligent? That it was conscious? Before you answer, think about whether you believe that other people around you are conscious, and how you arrived at that conclusion. You may also find the next chapter in The Mind's I interesting. It is a dialogue written by Douglas Hofstadter about the Turing test. There is a chapter in Churchland's book Matter and Consciousness describing research in Artificial Intelligence, or A.I., which has been very successful in some areas, but less so in others. A different view on the abilities of computers comes from the paper "Minds, Brains, and Programs" by the philosopher John Searle. It is chapter 22 in The Mind's I. Searle appeals to our intuitions to argue that no computer will ever be able to understand as we do. |
If our high level cognition operates according to a set of rigid rules, like a machine, then we ought to be able to reproduce those processes by building those rules into a computer program. Concocting such programs has been the goal of a field of research called Artificial Intelligence. Some researchers in Artificial Intelligence have tried to reproduce our mental processes in machines. How do we know whether the machine thinks the same way we do? That question brings us to another question: How do we know whether othe people think the same way we do?
If not, what are they missing?If so, how will we know when they are conscious?
Has anyone ever built a machine that is conscious?Could anyone ever build a machine that is conscious?
How can we decide if computers are capable of thinking, as we are?Turing proposes a test: Can the computer do things that we would interpret as thinking if we saw another person doing them?
Note: The Turing Test was proposed as a test of thinking, not consciousness.
- A judge carries on two conversations via keyboard and screen.
- One is with another person.
- The other is with a computer.
- Can the judge determine which is which?
Searle rejects the Turing Test
Focuses on "understanding"Even if a computer were programmed to behave just as we do, it would not be able to understand in the way that we do.
Searle's Chinese Room Argument
- One person locked in room; does not understand Chinese.
- Chinese speakers write sentences in Chinese and pass them into room.
- Person inside uses complex book of rules to produce responses in Chinese.
- The Chinese speakers believe that they are having a conversation with someone who understands.
- The person inside understands nothing about the conversation.
- They are like a computer that generates responses without understanding.
Searle claims that our thoughts have meaning (or intentionality), but the symbol processing in the computer does not. According to him, the computer has syntax, but not semantics.
Is Searle correct? Are computers incapable of understanding? If the person inside the room cannot understand the conversation, is it possible that the entire system understands?
If we can understand and computers cannot, what do we have that they do not? Is there something special about biological tissue that makes understanding possible? If so, what is it?
Searle's Chinese Room thought experiment is an appeal to your intuition about where consciousness arises and where it does not.
How useful are our intuitions on these issues?
Searle refers repeatedly to the concept of intentionality. This term has special meaning in the philosophy of mind. Something has intentionality if it is about something else or refers to something else. Thus, most of our mental states have intentionality. If I intend to set my clock back tonight, my mental state is about my clock. Intentions are not the only states that can have intentionality, however. Beliefs and desires can also be about something. I can believe that my car is in the parking lot, and I can desire an ice cream cone, both of which would be intentional states. When Searle says that our brains have intentionality but computers cannot, he is saying that our thoughts can be about things outside our head, but the representations inside a computer cannot.
Note that in this debate, Searle is willing to assume that we can write a set of rules that produce behavior exactly like that of a human. The debate is not over whether such a set of rules can be created, but whether a system operating under such rules would be capable of the same sort of understanding that we do.
Be sure to think about the concept of understanding, and what it means to understand.
Marvin Minsky compared the concept of Understanding to the concept of Life. Life used to be a very mysterious thing, but modern biology and biochemistry have provided us with a pretty good understanding of the mechanisms underlying life. If we take apart a living thing, looking for the life inside of it, all we see are biochemical mechanisms. If we look inside a system that is capable of understanding, should we expect to find components that exhibit understanding themselves, or merely pieces of a mechanism?
If you are interested in recent attempts to pass the Turing Test, here are some Web sites to explore.
Next class: Unconscious Processing
Psych 391D:
Consciousness 
Kyle Cave
Psychology Dept.
U.
Mass.
