main page




Schools of Thought in A.I. (the First 50 Years)






0 Outline of Topics
  AI - The Field, the Idea  
  Schools of Thought in AI  
  Classical AI / "good old-fashioned AI" (GOFAI)  
  Behavior-based robotics  
  Missing in Action  













Artificial Intelligence - The Field, the Idea
  Artificial Intelligence / Gervigreind "Artificial" as in "artificial flower", i.e. man-made
Gervigreind, eins og gerviefnin
  Basic idea An engineering approach to the phenomenon of intelligence
  Early Founding Fathers Rene Descartes, Charles Babbage, Alan Turing
  Dartmouth Conference 1956 Considered official founding of the field, by John McCarthy, Marvin Minsky and others.
  Scientific Foundation Computer Science - Intelligence can be formulated as an information transmission and transformation task
  Truly Interdisciplinary Intersects Lingustics, Psychology, Neuro- and Brain Science, Philosophy (epistemology, philosophy of mind)
  Related field Cognitive Science: The computational study of natural intelligence
  Alan Turing

Legacy: Colossus, Computability, "The Turing Test"

Paper: Computing Machinery and Intelligence (1950)

















Schools of thought in AI
  Cybernetics Starting in the 1940s. Example works: Norbert Wiener's 1948 "Cybernetics, or Control and Communication in the Animal and Machine" and continued work such as WHAT THE FROG'S EYE TELLS THE FROG'S BRAIN by Lettvin, Maturana, McCullochs & Pitts, 1968.
  • Effort to merge engineering/computer science and biology into a new field
  Classical AI / "good old-fashioned AI" (GOFAI) Starting 1956: The Dartmouth Conference
  • Logic
  • Games
  • Isolated models
  • advanced competences
  • static structures/knowledge
  Behavior-based robotics Starting around 1985 with R. Brooks' work on subsumption architecture
  • reactivity, speed
  • architectural simplicity
  • Holistic models of perception and action
  • Robotics
  • Simplicity
  Cognitive Science An effort to integrate knowledge from psychology, neurophysiology, engineering, computer science, philosophy, lingustics in the study of mind. Initially pushed forward by Carnegie-Mellon. Possibly what the Cybernetics movement intended to become.














Classical AI

"Classical AI": Top-down approach

Term used to describe "reflective" or "deliberative" approaches to AI


Example typical topic (i.e. caricature)

  • Chess
  • Argument:
    • No animal except humans can play chess
    • Everyone agrees that humans have "general intelligence"
    • Thus, a machine that can play chess must have general intelligence
    • Result: A lot of software that can play chess, yet have no general intelligence
  Contrast with "Behavior-Based AI"

"Classical AI": Sometimes it means:

  • everything but behavior-based AI

and somtimes it means:

  • most of what was done in AI in the first three decades, and anything that follows those traditions
  Classical AI: Topics
  • Knowledge: Correctness, completeness
  • Making models of the world to base decisions and plans on
  • Representing knowledge to support simulations of high-level thinking and knowledge
  • Specialized and general knowledge
  • Human communication and natural language

"Top-down" approach

  Classical AI: What it's good for
  • Inferencing
  • Games (search, planning)
  • Expert knowledge representation and use
  • Symbolic/semantic representation
  • Natural language
  • Rule-based systems
  Missing pieces
  • Artificial hearing
  • Emotion
  • Preferences
  • object vision
  • Artistic expression
  • Innovation and creativity














Behavior-Based AI

"Behavior-based AI"

Term used to describe "reactive" architectures with a tight coupling of sensors and actuators


Example typical topic (i.e. a caricature)

Autonomous vaccum cleaner

  Contrast with "Classical AI"

"Behavior-based AI": Sometimes it means:

  • everything that connects sensors directly to actuators

and somtimes it means:

  • most of what was done in AI in response to the failure of "classical AI" to come up with the goods
  Behavior-based AI: Topics
  • Complete systems
  • Perception-action relationship
  • Robotics
  • Situated systems
  • Embodied systems
  • Behavior and architecture, as opposed to knowledge

"Bottom-up" approach

  Behavior-based AI: What it's good for
  • Insect-like behaviors
  • Reactive planning
  • Instinctive behaviors
  • Robust simple behaviors in complex, realtime environments
  • As a robust basic component in a larger system

Ronald Arkin: Behavior-Based Robotics, MIT Press


Java simulator
(Braitenberg, Subsumption, more)















Missing in Action
  In classical AI
  • Breadth; holistic architectures for agents
  • Reactivity and handling of real-time
  • Robustness
  • Graceful degradation
  • no symbolic grounding
  Behavior-based AI
  • Reasoning
  • Common sense knowledge
  • Prediction, long-term planning
  In both classical and behavior-based AI
  • Artificial hearing, smell
  • Emotion
  • Preferences
  • object vision
  • Artistic expression
  • Innovation and creativity
  Hybrid approaches Combine ideas from both classical and behavior-based AI, and often other schools of thought