My answer to: Exercise 4: Definitions, definitions Which definition of AI do you like best? How would you define AI?
Let's first scrutinize the following definitions that have been proposed earlier:
- "cool things that computers can't do"
- machines imitating intelligent human behavior
-
autonomous and adaptive systems Your task:
-
Do you think these are good definitions? Consider each of them in turn and try to come up with things that they get wrong - either things that you think should be counted as AI but aren't according to the definition, or vice versa. Explain your answers by a few sentences per item (so just saying that all the definitions look good or bad isn't enough).
- Also come up with your own, improved definition that solves some of the problems that you have identified with the above candidates. Explain with a few sentences how your definition may be better than the above ones. Please read the above instructions carefully and answer both of the items above in the text box below. Your answer will be reviewed by other users and by the instructors. Please answer in English, and check your answer before clicking 'Submit' because once submitted, you can no longer edit your answer.
"cool things that computers can't do" This definition captures the idea that AI is always on the horizon, the next currently unsolved problem. While the definition provides direction for the future, the past can be difficult to summarize since classic AI programs are no longer categorized as AI to the present. The definition is non-static and in continuing flux.
machines imitating intelligent human behavior This definition captures the black box behaviorist approach to AI. Rather than dealing with elusive concepts like mind and thought, mechanically reproducing intelligent behavior is a clearly measurable standard. However, imitations aren't always equal to what they imitate. An AI program behaving intelligently may not actually be intelligent even if we treat it as such. This definition lacks a model of the program's internal processing.
autonomous and adaptive systems This definition captures essential aspects of our expectations of AI programs, to be self guiding and able to learn over time. However, it's broad reach causes vagueness and ambiguity. By these terms we can quickly misapply words like narrow and general intelligence or assuming understanding on the part of the program.
-
Artificial intelligence is an autonomous adaptive system which behaves rationally to solve a problem. I like this definition because AI programs are inherently tied to the problem which it's trying to solve. Unlike saying things computers can't do, we can discuss AI programs in relation to problems whether they are solved or unsolved. The behavior is the actions to solve the problem, and with rationality we can discuss optimal paths and the internal processes of the program. Grounding with the problem helps to narrow what is meant by autonomous and adaptive to condense these "suitcase words."