Overall Reading | |
---|---|
Brookshear: | Ch. 10.1-10.3, 10.6-10.7 |
Decker/Hirshfield: | Mod. 9 |
Outline:
Further work came out of a 1956 workshop at Dartmouth College sponsored by John McCarthy. In the proposal for this workshop, McCarthy used the phrase a "study of artificial intelligence."
Three views of research within AI:
If computers are ever so good at participating in coversations that you can not distinguish, then Turing says that the computer has displayed intelligence
Curiously, if someone claims to have a machine which possesses intelligence and consciousness, how can they prove it? How can you disprove the claim?
We credit other humans with intelligence and consciousness, not because we are sure that they have thoughts and emotions, but because they behave as if they do.
Can we produce a legitimate model of the workings of the human brain? We seem to know that the brain is made up of many neurons which are interconnected. Each individual neuron has a reasonably simple behavior. It is the complex combination of them all that we don't understand.
A comparison:
That is, can we figure out how the human brain works. At basic level, individual neurons are rather simple.
When you look out a window and see a tree, did you reason to figure out you are seeing a tree, or do you just know it?
"Thinking about this topic is like pulling teeth"
"Three wolves and three chickens have to cross a river using a boat which can only hold two animals at a time..."
We will look at three distinct levels of ability:
Let's walk through the issues for each individual ability.
What are the minimum requirements for understanding language?
Must understand vocabulary.
Must understand grammer (syntax)
But we use knowledge far beyond that. There are many potential ambiguities in natural languages, due to:
Examples:
Who was the person who later went to the store?
Is the word 'racing' a verb or an adjective? Is the sentance describing what kind of horses they are, or is it telling what some people are doing?
Even if you know the part of speech, certain words have several meanings. Which one is reasonable interpretation?
Examples:
'lies' is a verb. But does it mean to recline or to deceive?
Well, it seems unlikely that someone can tell a fib while sleeping, so I probably assume that this meant Ron was reclined.
Examples:
Do you interpret this differently if you hear it in a restaurant versus in an aquarium?
Does this have a different meaning if said by a stranger on the street than from your boss, after you walk into a meeting 20 minutes late?
Whose hand? Are we talking about a baseball player or a zoo keeper?
Did she have a round object, a good time, or a formal dance? Perhaps reading more of the story would help us in understanding.
Examples:
Let's assume we already know that this is asked by a stranger on the street. Do you think the stranger is expecting you to answer, "Yes, I do."
Almost certainly not. This is not really a question. It is informative.
Was I simply expression my feelings or was there a question/request in this sentance?
Examples:
Fortunately, I happen to know that he placed his paint on canvas (as opposed to skin).
So why was the waitress upset? Because it is customary to leave a tip in addition to paying the check. Of course we need to have knowledge about this real-world custom to be able to understand. No part of the text is going to make this explicit.
Certainly cannot just translate words:
"The spirit is willing, but the flesh is weak."
[translated to Russiona equivalent of]
"The vodka is acceptable, but the meat has spoiled."
Besides the issue of interpretting text, you need to convert the sound to the words. But how do you know where one word ends and the next starts? How do you determine the words, which are not always spoken in the identical way?
People use clues such as inflections, context, expectations.
How do you keep track of huge amounts of knowledge and then use it to
draw inferences.
Keep track of facts; draw inferences.
(The Decker/Hirschfield InferenceEngine)
Keep in mind that human intelligence is tough to beat.
(Example from Monty
Python)
Here is the HolyGrail logic, as a datafile which you should save to
disk, and then load from within the
InferenceEngine.
Repeated experiements with feedback (both positive and negative).
Adjust parameters.
Example: Robot throwing darts.