What is AI? - TAMU Computer Science Faculty Pages

Download Report

Transcript What is AI? - TAMU Computer Science Faculty Pages

• What is Artificial Intelligence?
– not programming in LISP or Prolog (!)
– depends on your perspective...
• a method for modeling intelligence
• a method for studying human cognition
• a method for building complex programs
• Answer from a Philosopher: a method for
modeling intelligence
– How do you define “intelligence”?
•
•
•
•
•
•
ideal: syllogisms...laws of thought...logic
need for embodiment? mind-body duality (Descartes)
physical brain required? Chinese Room experiment
Symbol Systems Hypothesis (Simon and Newell)
grounding, mechanization, novelty, adaptiveness,
animals? souls? free will?
– operational definition: acting intelligent
• Turing test; what is the correct measuring stick?
– no contact, response characteristics, news, humor, gender...
• Eliza, chatter bots, Loebner prize, Deep Blue
• Answer from a Psychologist: a method for studying
human cognition
– strengths:
•
•
•
•
perception, ambiguity, interpretation
ill-formed problems; weighing multiple criteria
judgement, common-sense, expertise
“insight”, analogy, Eureka effect
– weaknesses:
• calculations
• limited, selective, and maleable memory
• confirmation bias; role of emotions?
– behaviorism vs. info-processing metaphor (I/O, internal rep)
– connection to language
• Sapir-Whorf hypothesis (verbal representations)
• concepts, intension/meaning, maps, skills, automation
• Answer from an Engineer: a method for
building complex programs
– need more than just C++ or java (or OOP)
– search algorithms, inference techniques,
methods for dealing with uncertainty...
– knowledge-based programming
•
•
•
•
Feigenbaum: “In the knowledge is the power”
high-level programming languages, expressiveness
expert systems, logic, McCarthy
real-world knowledge bases, ontologies, Cyc
– intelligent agents: decision-making
– control theory, “cybernetics,” Kalman filters
– decision theory: Bayes, Markov
• Relationships of AI to other fields:
– Economics
• rational decision-making; satisficing (Simon)
– Mathematics
• computability; Godel’s Incompleteness Theorem;
logic and number theory; Leibnitz
– Neuroscience
• neural networks, Minsky
• connectionism, distributed representations,
grounding