Artificial Intelligence

This is a course about AI fundamental concept, learning models and relative techniques.

Definition

The definition of AI can be classified into four dimensions.

Thinking humanly	Thinking rationally
Acting humanly	Acting rationally

• Acting humanly Turing Test
  Turing Test testing whether machines have intelligent behavior like human. A computer wanted to pass the test should having the capabilities like: Natural language processing, Knowledge representation, Automated reasoning and Machine learning.

  • Category

    • (Normal) Turing Test: Avoiding direct physical interaction between interrogator and computer.
    • Total Turing Test: Providing computer vision and robotics.

  • Category about AI

    • Strong AI: Can be conscious.
    • Weak AI: Conscious is not compulsory.

• Thinking humanly The Cognitive Modelling Approach（认知建模法）
  To find out how human think, we have the methods like: Introspection（内省）and Psychological experiments. Analyse and use it to determine whether a program of mechanism can thinking like human. Setting up a new filed called Cognitive Science which is a combination of psychology and AI.

  • General Problem Solver (GPS)
    Focusing on reasoning. Checking whether the reasoning steps are match with human's.

• Thinking Rationally The Law of Thought Approach（法则思维法）
  Right thinking.

  • Obstacles
    1. No 100% certain knowledge.
    2. Solving problems "in principle" and "in practice" is difficult. (E.g. Having a theory but it can not be done in a reasonable time.)

• Acting Rationally The Rational Agent Approach（理性代理法）
  Consider AI as the study and construction of rational agents.
  *Agents is something can perceive and act.

  • Advantages (compare with The Laws of Thought Approach)
    1. More general
    2. More amenable to scientific development

References

[1] Yiu Ming Cheung. COMP 7015 Artificial Intelligence. Hong Kong Baptist University, 2020.
[2] Turing test - Wikipedia (opens new window).