Allan Turing proposed the “Imitation Game” to circumvent the question “Can machines think?” by providing a more precise AI goal. It involves a machine and a human, locked in a room, conversing with a human interrogator who aims to identify which entity is a machine while both human and machine convince the interrogator they are human. The machine passes the intelligence test when the machine’s responses and the human are indistinguishable to the examiner. It implies that a computer could only imitate a human being if it thought like one. In this essay, I will argue the conditions for intelligence represents by Turing Test are sufficient; hence the test is a reliable indicator of intelligence.
According to Lawhead (94), intelligence is an object’s capacity to respond to the changes in the requirement effectively. It is the process of using the available information to solve a problem or perform a task. Thus, intelligence in humans is the ability to understand and operate on knowledge stored in their minds. Mechanisms to represent the environment, change and update our models, and implement change are essential intelligence components. Physical environments affect cognition since human beliefs and thoughts are grounded in their environments. Thus, a machine could be considered intelligent if it effectively and efficiently uses information from the environment and uses it to perform a task or solve a problem. On the other hand, Turing Test’s criterion for intelligence is making a decision solely in external and observable characteristics. Thus, intelligence can be ascertained through observations of an entity’s interactions with the environment without knowing the composition or origin of the intelligence.
Machine state functionalists assert that a machine whose operation can be fully specified by a set of instructions can be regarded as a creature with a mind (Putnam, 36). The program acts as a mind, where for each state and set of inputs, the program specifies the machine’s probability to enter some subsequent state and produce some results. Hence, by observing a machine’s interaction with the environment through its relation to input, output, and one another, it is possible to understand and reproduce intelligence in an abstract computational device. However, one major issue that arises is the lack of real-world experience by machines to possess human intelligence since intelligence is a physical world factor. A machine cannot experience input from the real world the same way humans experience it, limiting its intelligence.
However, tying general intelligence to only human experience undermines the Turing Test’s reliability to test intelligence. A machine’s failure to understand the relationship between an input and the world does not necessarily imply that it should be labeled unintelligent (Savova and Leonid, 2). The machine’s different embodiment, design limitations, and experiences create its difference with a human. Of course, a person does not qualify as intelligent if their verbal behavior does not have accompanying knowledge. However, the assertion is wrong if we consider that a blind or deaf person can be considered intelligent as other humans despite their lack of knowledge in some aspects of the real world. Although their inability to relate to the real world and visual- or oral-based is accidental in this scenario, it does not bear on their innate intelligence. Thus, the Turing Test of intelligence is reliable since some humans are intelligent despite being deprived of real-world experiences.
One of the Turing Test’s significant aspects is the interaction of the interrogator and the machine only through language and the prohibition of physical or visual contact between the two. While this interaction method seems fair, machines should not be labeled unintelligent based on what makes a human intelligent. Language is crucial to recognizing human intelligence since it is not logical to label a person before some interaction with them. Humans interact through language to determine whether a human is intelligent. So a human can ascribe intelligence to other humans; hence they are human. Therefore, if language captures the human intelligence Turing Test is a reliable intelligence test since machine and human responses are indistinguishable.
According to Premack, pg.319, language usually captures most, if not all, intelligence, even sub-articulate thoughts. However, artificial intelligence requires duplicating human cognitive skills to learn sub-articulate thoughts, which is the type of cognitive operations humans are consciously unaware of and intelligently articulate. An example of sub-articulate thought is the plural of imaginary English words that do not follow the conventional pluralization rules (Michie, 9). However, with the development of machine learning and the internet, machines can learn the cognitive skills of a human. Thus, once a machine learns to duplicate human sub-cognitive activities using datasets from the internet and machine learning technique Turing Test is a reliable intelligence test.
Turing Test provides sufficient conditions to ascertain intelligence considering Turing’s intelligence criterion, where the decision is based on external and observable characteristics. The definition of intelligence involves using cognitive abilities to adjust to the environment, which is influenced by human experiences. However, relating inputs to real-world experiences is not conditional on being considered intelligent since there is deaf or blind human, who are considered intelligent. Also, language is an essential factor in recognizing human intelligence since it articulates human cognitive and sub-cognitive activities. Since the machine’s and human’s responses are indistinguishable in Turing Test, which is solely based on language interaction, it is a reliable intelligence test. However, the machine should master human cognitive activities to articulate intelligence.
Lawhead, William F. Cengage Advantage Series: Voyage of Discovery: A Historical Introduction to Philosophy. Cengage Learning, 2014.
Michie, Donald. “Turing’s test and conscious thought.” Artificial Intelligence 60.1 (1993): 1-22.
Premack, David. “Is language the key to human intelligence?.” Science 303.5656 (2004): 318-320.
Putnam, Hilary. Brains and behavior. Blackwell, 1968.
Savova, Virginia, and Leonid Peshkin. “Is the Turing test good enough? The fallacy of resource-unbounded intelligence.” IJCAI. 2007.