Intelligent Systems And Their Societies Walter Fritz

Learning with no prior knowledge
of the environment

 

The "General Learner" is a system that has as inputs elementary sensations and outputs elementary actions. It has no programmed knowledge about its environment. But we have programmed the ability to learn. It learns exclusively from its inputs, outputs, their effects on the environment and approval or disapproval by a person (like a child does). It creates "concepts" representing things in its environment and creates response rules which indicate what to do in a given situation (response rules are expressed with concepts).

The "General Learner" just relates inputs to outputs. It does not know anything about what these inputs and outputs mean, what they stand for. We have programmed only how to learn, that is how to relate the system's input with its output.

If you think about it, our brain can also never know what the patters of neuron impulses it receives, stand for. Then it sends out other patterns of impulses, which it has found useful, without any chance of knowing what elementary actions these provoke.

Normally the GL learns by just observing the actions of persons, as when two persons are playing tic tac toe or nim, and storing this experience as a response rule. Then the GL plays, using this experience. When a person sees an incorrect response by the GL, she or he can signal disapproval by pressing the down arrow (the up arrow is used for indicating approval).

So you see that there is no knowledge of the environment at the start of the run, but much knowledge about how to create concepts, response rules and about how to generalize them.

 

For continuous reading, like a book - continue here
Jump to the General Learner document /Artificial IS document /e-book Contents.


Last Edited 11 April 2013 / Walter Fritz
Copyright © New Horizons Press