Most software specializes in a restricted domain

• Remarkable success in focused applications:
  • IBM's Deep Blue beat Chess Grand Master Garry Kasparov in a 6-game tournament in 1996
  • Grinnell's expert system for placing incoming students in mathematics and computer science courses (based on transcripts and standardized tests) does a better job than faculty, BUT
    faculty clearly are better after students arrive on campus and can be interviewed
  • Carnegie-Mellon's "NavLab group builds computer-controlled vehicles for automated and assisted driving." (Navlab Home Page.) We "have demonstrated systems capable of autonomously driving the Navlab vehicle for long distances at speeds of up to 90 mph." (Dean Pomerleau, http://www.ri.cmu.edu/people/pomerleau_dean.html, retrieved 10 January 1007. )
    
    
• Few programs try to cover many domains (e.g., medicine, tax preparation, transportation, aerospace, manufacturing)
• Success typically do NOT try to simulate processes of the human brain (although neural networks based on a rough, approximate model)

Computers can "learn", based on new situations and experiences

• Data may be added or changed in a database
• Processing may be adjusted (e.g., weights changed in neural networks)
• Examples:
  • Search engines/commercial Web sites may remember past interactions for a user to tailor interactions to apparent user preferences
  • Spam Assassin refines its definition of spam, based on input from users

created 8 January 2007 last revised 15 February 2007

previous   next