|Intelligent Systems And Their Societies||Walter Fritz|
For a number of years, computer scientists have written programs that are complete intelligent systems. Not all are identical to the "intelligent system" that we described in a previous chapter. Nevertheless all have an input corresponding to senses, a choice of actions based on response rules, sometimes called "productions", and the ability to act, be it as graphics on a computer screen, as a text output or as limb movements. Most have a memory for storing experiences and the ability to learn.
Sometimes the brain does not do the action immediately but uses its imagination. It selects a response rule and determines what situation results from the action. Then it selects again an action for this new situation and determines the probable result. Thus it can choose not only one response rule but a complete plan of action.
Why build them?
Possibly you ask yourself: Why does anybody want to build complete Artificial Intelligent Systems? Well, they could free us from work and provide a much higher standard of living.For details see here (for continuous reading, like a book)
Do They Really Think?
It is often said that computer programs, of the IS type, do not think, they are only simulating or modeling thought. Let's look at this as follows: We have no problems saying that birds fly. Likewise, we say that airplanes and even model airplanes fly. Yet, we say that a plane in a flight simulator program on a computer does not fly. Why is this?
The concept of flying originated with birds, we say that "a bird flies". Also we say that both full size and model airplanes fly because we see that they comply with all the main functions and the behavior of flying. But a plane on the screen of a computer does not fly because it is only a picture, many functions or actions of flying do not exist. There is no lifting sensation, there is no wind around the wings which supports the plane, so we say that it does not fly.
So what about artificial intelligent systems? We have to judge each IS as a separate case. If they comply with the main functions and processes of the brain, then we can say that they really think. We believe that this is the case, for instance, in the General Learner 3 (For continuous reading, like a book - do not enter here now) and some others.
It is not enough to just write a computer program that is an intelligent system and runs within a computer with an output on a screen. We need to build complete systems that act in our human environment. It seems, that to be useful, such a program should have an intelligence somewhat like a human one. And for that it needs human like senses and limbs, so it can have similar experiences and create concepts for actions somewhat similar to our human ones.
The page on symbol grounding(Enter for continuous reading, like a book) explains how the computer concepts are based on sense information.
At present the development of humanoid robots is pretty fast:
Biologically inspired computer Architectures: BICA-1009
Brain inspired cognitive systems: BICS-2010
Artificial General Intelligence: AGI-1009
Association for the Advancement of Artificial Intelligence: AAAI-10
International Conference on Epigenetic Robotics (Self constructing robot brains)
International Conference on Robotics and Automation: IEEE ICRA 2009
International Symposium on Humanoid Robots.
Some interesting existing robots are:
There are yearly competitions between robots, for instance:
Also there is a site, that shows over a hundred mayor robot projects existing now (2010): Android World (Exterior link).
In most of the robots shown above all actions that they can do, are programmed in detail.If you have a short list of required actions, those can then be programmed in a short time. But the robot cannot do anything else. If you want a robot that can do many things, then it is better that it learns them, like a person does. Programming would take too long. In other words, in this case we should only program the learning capability.
It looks to me, that Piaget's theory of cognitive development (in a child) should be taken into account when developing a cognitive architecture that can learn. For information on this theory see:Theory of cognitive development(Exterior link).
The Nao humanoid robot from Aldebaran robotics(Exterior link), is using Piaget's theory of cognitive development.
Human and animal brains work, using neurons. Neurons have a certain input from other neurons and a certain output to other neurons. This chain starts with sense impressions, that excite the first neurons and ends with the last neurons activating some muscles. In biology this whole process is called stimulus and response. It seems that even the most abstract thoughts occur using this process. Here the last step is exciting the muscles that produce (silent or loud) speech.
It looks to me that an artificial brain should imitate this process.
Some interesting robot brains
Here we indicate some interesting robot brains (also called cognitive architectures), that learn what actions to do, instead of having them programmed.
Also see Wikipedia:
For continuous reading, like a book - continuehere.
Jump to Contents / top of this page
Copyright © New Horizons Press