The most basic cognitive construct that comes to mind is the Chinese Room. This idea was created to argue against Strong AI and it was mentioned by Dennett. I think we can get some more utility out of it.
The basic notion was that you have a room where pieces of paper with questions are fed through a slot. The person in the room looks up the question in a book and copies the answer on a sheet of paper and feeds this back out through another slot. To make the point that the answers had to come from the book it is stated that the questions and books are in Chinese and that the person in the room doesn't speak Chinese. The questions are looked up strictly by matching the patterns of the Chinese characters.
This construct does not require a specific language or that books are used or pieces of paper. The facilitator in the room could be a person but could also be computer. These details are not important. Some will probably note that this is related to the Turing Test. The method of communication could be writing or typed or voice. The answers could be in books or on punch cards or on tape or on large arrays of hard-drives.
We agree that the neither the data nor the storage medium are conscious. We agree that the facilitator provides no information. So, we start with no consciousness. Watson is a modern example of this.
Questions are entered but as they become more complex the data will not have matches. So, we add more data. This works for a little while however we then realize that it will be impossible to have all permutations of reality. The number of questions is uncountable in the same way that real numbers are uncountable. Therefore there isn't enough information available on the internet and in the Library of Congress to answer these questions and there would never be no matter how large we made the data store.
Assertion 1.) It is not possible to create a working Chinese Room based solely on pattern matching.
If Watson can't do this task on even a theoretical level then we need to add something. Let's try using formal logic and formalized data. This should let us make additional deductions and allow us to answer more questions. Cyc is a modern example of this.
So, we assume that our data is in sets that show relationships. For example, a Cocker Spaniel is a dog. A dog is a mammal. A mammal is a living creature.
This would allow us to generalize. For example, if we have a question about Rex who is a dog we can assume that anything that is generally true of a dog is also true about Rex such as having four legs and being warm-blooded. We would also know that Rex is not a cat or a tree. This is a big improvement. We could state, for example, that John is married to Linda and then ask who John's wife is. This system could come up with the right answer based on marriage as a set and husbands and wives as subsets whereas simple pattern matching would not.
Any simple relationship should be definable in terms of set theory and logic. But, suppose we asked why a helium balloon will rise if you let go of the string. This system would not provide an answer.
Assertion 2.) It is not possible to create a working Chinese Room based on pattern matching, set theory, and logic.
This seems to be a good start and then we can continue.
The basic notion was that you have a room where pieces of paper with questions are fed through a slot. The person in the room looks up the question in a book and copies the answer on a sheet of paper and feeds this back out through another slot. To make the point that the answers had to come from the book it is stated that the questions and books are in Chinese and that the person in the room doesn't speak Chinese. The questions are looked up strictly by matching the patterns of the Chinese characters.
This construct does not require a specific language or that books are used or pieces of paper. The facilitator in the room could be a person but could also be computer. These details are not important. Some will probably note that this is related to the Turing Test. The method of communication could be writing or typed or voice. The answers could be in books or on punch cards or on tape or on large arrays of hard-drives.
We agree that the neither the data nor the storage medium are conscious. We agree that the facilitator provides no information. So, we start with no consciousness. Watson is a modern example of this.
Questions are entered but as they become more complex the data will not have matches. So, we add more data. This works for a little while however we then realize that it will be impossible to have all permutations of reality. The number of questions is uncountable in the same way that real numbers are uncountable. Therefore there isn't enough information available on the internet and in the Library of Congress to answer these questions and there would never be no matter how large we made the data store.
Assertion 1.) It is not possible to create a working Chinese Room based solely on pattern matching.
If Watson can't do this task on even a theoretical level then we need to add something. Let's try using formal logic and formalized data. This should let us make additional deductions and allow us to answer more questions. Cyc is a modern example of this.
So, we assume that our data is in sets that show relationships. For example, a Cocker Spaniel is a dog. A dog is a mammal. A mammal is a living creature.
This would allow us to generalize. For example, if we have a question about Rex who is a dog we can assume that anything that is generally true of a dog is also true about Rex such as having four legs and being warm-blooded. We would also know that Rex is not a cat or a tree. This is a big improvement. We could state, for example, that John is married to Linda and then ask who John's wife is. This system could come up with the right answer based on marriage as a set and husbands and wives as subsets whereas simple pattern matching would not.
Any simple relationship should be definable in terms of set theory and logic. But, suppose we asked why a helium balloon will rise if you let go of the string. This system would not provide an answer.
Assertion 2.) It is not possible to create a working Chinese Room based on pattern matching, set theory, and logic.
This seems to be a good start and then we can continue.
via International Skeptics Forum http://ift.tt/1WAoLPD
Aucun commentaire:
Enregistrer un commentaire