Defining Artificial Intelligence
Learning Objectives
- You know what the term intelligence refers to, and you know of some definitions of artificial intelligence.
- You know of challenges in defining artificial intelligence, and you know of the AI effect.
- You know of the distinction between weak and strong AI.
Intelligence
As per the Merriam-Webster dictionary, intelligence refers to “the ability to learn or understand or to deal with new or trying situations” and to “the ability to apply knowledge to manipulate one’s environment or to think abstractly as measured by objective criteria (such as tests)”.
That is, intelligence encompasses the ability to learn, understand, and act in new situations, and the ability to draw from existing knowledge when doing so. Classically, intelligence has been assessed using a range of tests that measure cognitive abilities such as reasoning, problem-solving, understanding natural language, perceiving visual information, and so on.
As an example of what an intelligence test could look like, Figure 1 presents an question from Raven’s Progressive Matrices, which is a generic intelligence test. In the test, the participant would be presented with questions with a pattern of shapes and would be asked to identify the missing piece that completes the pattern in each question.
Author: Life of Riley (CC BY-SA 3.0).
Although intelligence tests have been critiqued, they have been and still are used as a way to measure cognitive abilities and to compare individuals’ with one another. There are also other types of intelligence, including emotional intelligence, which refers to the ability to recognize, understand, and handle one’s own and others’ emotions.
When discussing intelligence, it is also meaningful to distinguish intelligence from wisdom which relates to insight, judgment, and accumulated philosophical knowledge, and the ability to act in a sensible manner. Wisdom is often associated with experience and the ability to apply knowledge in a way that benefits oneself and others. In general, it is possible to be intelligent, but lack wisdom, and vice versa.
One could, for example, imagine an individual, who is intelligent in terms of being able to act in new situations, drawing from existing knowledge. However, if the scope of knowledge would be narrow, or if the individual would have been misinformed about the world, the individual might be limited in how they act in new situations, or fail to act in a way that benefits oneself or others.
Artificial intelligence
The term artificial intelligence, again as per the Merriam-Webster dictionary, refers to “the capability of computer systems or algorithms to imitate intelligent human behavior”. With the definition, the term encompasses the idea of creating systems or algorithms that can perform tasks that typically require human intelligence, such as reasoning, problem-solving, understanding natural language, perceiving visual information, recognizing speech, and making decisions.
To illustrate the extent of the definition, let’s study tic-tac-toe, where players put crosses and noughts in a 3x3 grid. Figure 2 presents an example of a tic-tac-toe game. In tic-tac-toe, each player puts one of their marks (cross or nought) to an empty spot in the grid. Finally, the player who gets three of their marks in a row (horizontally, vertically, or diagonally) wins the game.
Author: Mazeo.
The following writeup includes an algorithm — a sequence of instructions — for playing tic-tac-toe. The game would start with an empty 3x3 grid. If you wish, print out the algorithm on a piece of paper, and try playing the game with a friend or a family member so that you are following the instructions from the paper. Otherwise, study the algorithm and reflect on how it would fare in playing tic-tac-toe.
Say: Let's play tic-tac-toe. I'm X and will start.
Say: I put an X to a corner.
Act: Choose 1 corner from the game, and put an X there.
Remember the corner as "Move 1".
Say: Your turn, put an O to the board.
Act: Wait until your friend has put an O.
Say: Nice move!
Act: If your friend put an O to the opposite corner as "Move 1",
put an X to a free corner. Otherwise, put an X to the
opposite corner as "Move 1".
Say: Your turn, put an O to the board.
Act: Wait until your friend has put an O.
Act: If the board has a line with two X and en empty spot,
put an X to the empty spot and say "I won". Stop playing.
Otherwise, if the board has a line with two O and an empty
spot, put an X to the empty spot.
Otherwise, put an X to a free corner.
Say: Your turn.
Act: Wait until your friend has put an O.
Act: If the board has a line with two X and en empty spot,
put an X to the empty spot and say "I won". Stop playing.
Otherwise, if the board has a line with two O and an empty
spot, put an X to the empty spot.
Otherwise, put an X to a free corner.
Say: Your turn.
Act: Wait until your friend has put an O.
Act: Find an empty spot from the board, and put an X there.
Say: Thank you for playing.The algorithm is designed so that the player who follows the instructions would win the game if the opponent does not make any mistakes. The algorithm is based on the idea of creating a line with three X marks, and blocking the opponent from creating a line with three O marks. The algorithm also includes a strategy for the first move, where the player puts an X to a corner, and then reacts to the opponent’s moves.
Do you think that the algorithm exhibits “artificial intelligence”? Why or why not?
The above example highlights that there are ways to create systems and algorithms that can perform tasks that typically require human intelligence. At the same time, it also highlights the challenge of drawing line between what is artificially intelligent, and what is not.
This challenge relates to a phenomenon known as the AI effect. The AI effect highlights that as soon as a system or algorithm works, it is no longer considered artificial intelligence. In the words of John McCarthy, one of the pioneers of artificial intelligence, “As soon as it works, no one calls it AI anymore”.
For additional details, read John McCarthy’s Obituary by Bertrand Meyer.
There are also other definitions for the term artificial intelligence. As an example, artificial intelligence has been defined as..
- “the science and engineering of making intelligent machines, especially intelligent computer programs” (John McCarthy)
- “the science of making machines do things that would require intelligence if done by men” (Marvin Minsky)
- “The capacity of computers or other machines to exhibit or simulate intelligent behaviour; the field of study concerned with this. In later use also: software used to perform tasks or produce output previously thought to require human intelligence, esp. by using machine learning to extrapolate from large collections of data.” (Oxford English Dictionary)
At times, perhaps due to the AI effect, the definition has been left vague, and artificial intelligence has been introduced through a range of examples, such as game-playing programs, problem-solving machines, question-answering machines, and so on.
Weak and strong AI
Artificial intelligence has been also categorized into weak AI (or narrow AI) and strong AI (or general AI). Weak AI systems are designed to perform a specific task, such as playing tic-tac-toe, recognizing speech, or translating text from one language to another. Strong AI, on the other hand, would have the ability to perform any task that a human can do, and possibly, even have a mind in the same sense as humans have minds.
So far, the advances in the field of artificial intelligence have been in the domain of weak AI systems, while the development of strong AI systems has remained a long-term goal for the field.
There exists also critique towards the idea of strong AI. As an example, the philosopher John Searle introduced the Chinese room argument in the 1980s to challenge the possibility that a computer executing a program could have an understanding or be conscious. The key idea in the argument is that a system that can process symbols and produce responses to them does not necessarily understand the symbols, and that understanding requires more than just processing symbols.
For additional discussion and an extension of the topic, see Open Ended Intelligence: The individuation of Intelligent Agents, which discusses understanding of general intelligence, and posits that one way towards general intelligence could be through interacting agents and their coordination.
While the majority of actual advances have been in the domain of weak AI, strong AI has been a popular topic in science fiction. As a classic example, the HAL 9000 computer in the 2001: A Space Odyssey movie had human-like intelligence and the ability to understand and act in complex situations.
Depiction of AI in fiction has varied good to evil, and from helpful to harmful, and naturally the AI systems have been designed to serve the plot of the story.
As an example, if you are familar with the Star Wars franchise, have you ever wondered why the R2-D2 android that clearly is a strong AI only could communicate with beeps and boops?