Limitations Of The GL System
Now let us look at the limitations of the GL (General Learner) system. First, the system has physical limits which are created by the amount and type of senses it possesses and by the elementary responses it can perform. It has only two senses (the input devices of a mouse and a keybord), and its elementary responses are limited to producing drawings and words on a screen. Obviously the system cannot show intelligence beyond these limits.
Further, there are important "mental" limits, those due to its program functions for processing concepts and response rules. For example, the GL system cannot:
As you can see from the above limitations, much more work is still needed to reach a satisfactory artificial intelligent system.
- React to movement on the screen, as in the predicting of the future position of objects, and interacting with them. (We have implemented the function, at present, only partially.)
- Because it is not needed in the present environment of mouse, keyboard and screen, it has no function for the reduction of massive input data, such as the input data that could come from an CCD eye or a microphone.
- Name lines, curves, geometric figures, etc. drawn by a person (but given the name, it does produce the figure).
- Produce drawings in relation to existing figures (above, to the right of ...).
- Learn to continue grouped rule patterns (2 3 5 2 3 5 ...). (But it can continue sequences of numbers and letters in simple arithmetic progression, like "3, 4, 6" or "9, 7, 5" or "d, g, j, m".)
- Learn concepts like "next," or "previous."
- Recognize the same stone position, but in a different place of the board. (It needs a better function for composite concepts.)
- "Try out" a series of responses in "its imagination." It does not, at present, have an internal representation; it does not make plans. (A previous program made plans, but here we are trying to teach it to make plans.)
- Indirection: Draw it bigger (the last figure drawn).
- Negation, such as: "do not draw a line" - "move all circles except one."
- Conditional requests, such as: "only if there is a circle, draw a line."
- Answer questions, after typing a story, such as: who, what, where, when, how, why or which.
- Also the system cannot learn activities like driving a car or flying a plane due to sense and actuator limitations. (In principle, it should be able to learn these activities once some additional functions and their corresponding senses and actuators are incorporated.)
- Even with additional functions, the GL system cannot do tasks normal to humans such as writing a letter or working in a travel agency or in a bank. This is because these tasks require general world knowledge that GL has not learned due to time and computer memory space constraints.
For continuous reading, like a book - continue
Jump to the top of this document /General Learner / Artificial IS / e-book Contents.