• Unit 1: What Is Artificial Intelligence?

    Defining the term "intelligence" is surprisingly difficult. Alan Turing was among the first computer scientists to suggest what is now called the Turing Test to determine whether some software is exhibiting intelligence. This test has evolved as technology has evolved, but it stands up surprisingly well in capturing the key capabilities associated with intelligence even today. This unit describes the details of the Turing Test and the Full Turing Test and illustrates the concepts with many examples. You will also be provided with an overview of how scientists of different disciplines approach "intelligence". The key question is whether intelligence is defined by something innate or important within the agent being evaluated or more about whether the agent delivers rationally more effective outcomes to users, regardless of how it works internally.

    Completing this unit should take you approximately 2 hours.

    • Upon successful completion of this unit, you will be able to:

      • explain the different definitions of intelligence, including the Turing test(s) for intelligence;
      • compare and contrast the four different orientations for AI work; and
      • explain the dominant AI orientation for contemporary applications.

    • 1.1: The Turing Test

      The Turing Test, proposed by Alan Turing in 1950, suggests that if a computer running a program can produce results for a problem indistinguishable from the results produced by human beings, that is an "intelligent" program. As you will learn, the importance of the test is that "intelligence" is a matter of whether the outcomes produced by programs are comparable to those produced by human beings rather than whether the computational processes and algorithms that produce the outcomes resemble the thinking of a human being. The basic Turing Test is often described as minimally requiring knowledge representation (understanding of the world), logical reasoning (ability to deduce new information from known information), learning new things from experience (what is called machine learning), and natural language processing (the judge types inputs in English or some other human language and sees outputs/answers in the same language).

    • 1.2: The Four Types of AI

      Similar to the Turing Test, modern AI focuses not on how systems THINK internally but on how they ACT to produce outcomes. The four strains of AI work have been systems that THINK humanly, systems that ACT like humans, systems that THINK rationally, and systems that ACT rationally. Of these four different strains, modern AI is dominated by systems that ACT rationally. "Rational" in this context means "producing verifiably better outcomes". We use GPS navigation in our cars not because we understand HOW the system works but because we can see rationally that its recommended guidance works – we get to our destinations!