Approaches to A.I.
Bottom-Up, Top-Down, Generation Robots

The word "robota" in the Czech language means forced labor.  With the rise of artificial intelligence, programming of robots to do menial tasks, for a busy individual in the 90's, is a desirable and worthwhile attempt.  Many approaches to developing a robot capable of doing chores and other tasks exist.  Each theory, according to the respective researcher, is the best approach for developing an artificially intelligent robot.   "The trick is that you have to use the brain to understand itself..."-Thomas Hayden   Bottom-Up

The bottom-up approach is being implemented by Rodney Brooks at MIT.  According to the approach, the machine will discover the world on its own, the way humans do.  Cog (Cognitive) is the name of the robot used in Brooks' research.  Cog, created in 1992, has two black and white video cameras for eyes and is able to hear and grasp objects.  According to Brooks, ambitions for Cog in the future include distinguishing sounds so that it may understand the difference between a human voice and noise, for example, a lawnmower.     Top-Down

On the other end of the spectrum is an approach employed by Douglas Lenat, CEO of Cycorp.  Lenat's robot is named Cyc (pronounced "psych").  Cyc, created in 1984, is programmed with millions of common-sense facts which it uses to put two and two together using an "Inference Engine".  According to Lenat, learning occurs from what is already known.  Some of these common-sense facts are:

  Generation Robots

Hans Moravec is enthused about the projects with Cog and Cyc.  However, he feels that what should be a multi-step process is being thrown into one step.  Moravec's plan is a four step procedure.  He describes the first step as First Generation Robots.  First Generation robots are the equivalency of a small lizard.  They are mobile and understand their environment.  Moravec projects to be at this stage by 2020.  The second step, Second Generation robots, will take a decade longer.  These robots are equivalent to small mammals and in a way will follow basic Skinnerian training.  A sort of checklist is kept on errors and progresses made by the robot.  If the robot has failed to put away the dishes on one or two occasions, then eventually it will refer to the checklist, notice the number of errors in this particular action and then discontinue the action or refine it.  The difference between the First and Second Generation robots is that the First Generation will never learn.  By 2030, Moravec foresees the Third Generation robot which is equivalent to a monkey.  Not only will it correct itself, it will do so mentally before physically.  Put another way, it will think before it acts.  Also it will have a psychological model and refer to itself as having a mental life.  It will also be able to interact with others (humans and animals) respecting their space in the environment.  These robots will also be able to analyze.  The Fourth Generation Robots will be very similar to humans.  It will be one step above the Third Generation because it will infer ideas.  What is crucial to this study is that as research becomes more involved, costs have dropped due to cheaper means of attaining the same goal.

 Home Background  Definitions Approaches


 Implications   Re-examine

Back to Top
Back to Psybersite

This document was created April 19, 1998 and last modified on Tuesday, March 11, 2014 at 17:34:10.
This document has been accessed 7,324 times.
Please send comments and suggestions to