Alan Turing proposed an imitation game that would be used as a test for artificial intelligence. The game saw a machine being declared artificially intelligent based on the timing and imitation capabilities. This paper reviews the argument of Robert French that criticizes Turing’s imitation game in the paper titled “Subcognition and the Limits of the Turing Test” against that of Dale Jacquette entitled “Who's Afraid of the Turing Test?
Turing’s Intelligence Test’s Criticism and Objection
Alan Turing proposed an imitation game that would be used as a test for artificial intelligence. In his game, Turing proposed that if a machine would be subjected to the Turing test for five minutes and be able to imitate the interrogator to about 70% success, then it may be regarded as artificially intelligent. In his proposal, Turing also perceived the fact that such a machine would be in existence by 2000. This was about sixty five years ago. The Turing test, therefore, had a claim of the possibility of such a machine ever existing. The test raises philosophical questions as to whether the interrogator is indeed intelligent since the success is dependent on the interrogators’ approval. This dependency challenges the conclusive ability of Turing test to be achievable by an artificial model. As for me, it is true that no artificial intelligence, according to Robert French’s argument, will pass the Turing intelligence test if the appropriate types of questions are asked.
Robert French achieves to criticize Alan Turing using his own proposal in his test. He launches this criticism by allocating a philosophical meaning and pragmatic reason to the Imitation Game of Turing. French claims that, Turing philosophically meant that the artificial intelligence of a machine would be justified by passing his test. Pragmatic reasoning can imply that it will not take much time before a machine of this nature is built. French is in agreement with the philosophical-based claim of the test, but he refutes the pragmatic side of it. He supports the fact that the philosophical claim provided a pure and new test for intelligence and was able to avoid the philosophical quagmire that addresses the mind-body problem. This translates it efficiently into an operational definition of intelligence, i.e. “whatever acts sufficiently intelligent is intelligent”.
French argues that the Turing intelligence test is not a useful test for real intelligence because of its pragmatic claim that probes the human’s deep and essential cognitive areas. This is the reason why the Turing test could be only passed by models which have had a similar experience of human beings in the world. Based on this factor, French argues that the Turing test guarantee culturally-oriented human intelligence and not intelligence as a whole. To explain this, French proposes the Seagull test for flight.
According to the Seagull flight test, a machine will be said to have passed the test if it is conclusively indistinguishable from the seagull from the philosophical perspective. This test proves that it is impossible to categorize any machine with jets as a flying machine even though it can actually fly. The test demonstrates further that it is very probable that only seagulls from the Nordic island can pass the test. The implications of this test are clear to the Turing test, where an artificial entity might be extremely intelligent, but failure to respond to the interrogator’s questions humanly would fail the test. This makes the Turing test to be rather a test of humanly practiced intelligence than test of intelligence generically. French identifies the inability for the Turing test to threat intelligence as a continuum by inclining intelligence to the human-related intelligence. The Tuning test does not consider intelligence as a variable along species lines and even within certain species. This human inclination of the Turing test for intelligence is one of the reasons why French considers it not to be a sufficient test for artificial intelligence.
French also claims that there is no clear distinction between cognitive and physical level of questions. He draws this claim from the fact that the cognitive and subcognitive levels are inseparable. It means there is no much ability when it comes to separating subcognitive-based and non-subcognitive-based questions from others. It is this argument that finds the questions in the Turing test to be subcognitive, making it a test for human intelligence and not for artificial intelligence. This argument discourages the use of the Turing test in testing for intelligence in general. This happens due to the fact that if it is used in this perspective, it leads to categorization of intelligence. French then proposed that the philosophers of artificial intelligence need probably a theory of intelligence rather than a test of it. The author designated any question that provides an unconscious recognition of an experience as subcognitive, which refers to the subconscious associative network of the human mind.
French postulates a set of questions based on current associative priming research. These questions demonstrate that humans have developed a unique sense of associations among concepts of varying strength. This characteristic enables human beings to use less time in deciding a word when primed by an associative term. For example, butter and bread are associated with each other more than dog and bread. The machine would definitely fail such a test since it has never experienced the world as a human. The Neologism Rating Game would also require a machine to have experienced the world as we have in order to be able to pass this test. French further argues that it is impossible to isolate the cognitive level from the physical one using the example of exchanging knees in individuals of varying eyes. The experiences of these individuals show variations hence supporting the author’s argument of the disassociation lack between physical and cognitive levels. This cognitive limitation of machines makes them totally unable to pass the Turing test. French further maintains that some cognitive substrate is necessary to intelligence indeed.
French’s school of thought is opposed by philosopher Dale Jacquette by rejecting the interpretation of the Imitation Game taken by French. Jacquette argues that the very explanation of Turing’s ideas by French is inaccurate and, therefore, invalidates his criticism. That is, “Turing nowhere makes the philosophical claim that French attributes to him: if machines pass the test, then they are necessarily intelligent. On the contrary, Turing maintains that the question about whether machines are able to pass the test should replace the question "Can machines think?"”. Jacquette continues to argue that Turing would deny the interpretation that French had on his test since he had clearly stated his question in the article.
Jacquette opposes the imitation claim by French that the machines that pass the Turing test are necessarily intelligent by proposing that the claim will have no meaning to Turing. The author also claims that Turing only sets the conditions to satisfy the machines’ intelligence and restricted himself to a novel idea of having a new linguistic usage in future.
Jacquette criticizes the ideas of French in the Seagull flying test as a philosophical complain on intelligence rather than the reason of doubt in Turing’s prediction of a novel information technology in the future. He further argues that the Seagull test only focus on the candidates that fail the test; however, the researcher does not consider the candidates that pass the test. He claims that “If a candidate is capable of flight (intelligent), then it is able to pass the Seagull test." . Jacquette further argues that French only limits himself to intelligent entities that cannot pass the Turing test rather than considers entities that are very intelligent to pass the test as it was the objective of Turing
Jacquitte criticizes the fact that French limited the imitation game to have human interrogators and claim that Turing himself did not include the interrogators in his test. He implies that the idea of Turing was about the interrogators being human or nonhuman. He further explains that if a nonhuman interrogator is used in the Turing test, then the objection by French is invalid
Jacquitte maintains the argument that Turing’s question of a machine winning the imitation game remains unchanged as a sufficient condition for intelligence even if machines fail to pass the test. He claims that the failure of some machines to attain these conditions, even though they are intelligent, does not change the philosophical claim that Turing proposed.
I agree with Robert French’s argument that no artificial intelligence will pass the Turing intelligence test if the appropriate types of questions are not asked. In review of the French’s discussion on the imitation game proposed by Alan Turing, I find that the test was not specifically used for intelligence as a whole, but the artificial intelligence was tested in comparison to human intelligence instead. Due to the limited description of the parameters that Turing intended to use, he left a gap that would be used against his test. The fact that Turing did not explain who the interrogators were to be leaves a chance to assume that they were human. In return, it shifts his objective from general intelligence to human intelligence. Turing also failed to consider the diversity of intelligence along the species lines hence limiting the validation of the test to be used for artificial intelligence. I find Jacquette to be in agreement with French. It is evident in his statement: “French is correct to maintain that strong philosophical claim about the intelligence of machines that can pass the Turing test is standardly imputed to Turing’s discussion”.
French Robert proposes that there is no model of artificial intelligence being able to pass the Tuning’s imitation game because the test is inclined to human intelligence and limits the ability of any machine that has not experienced the life of a human being in the world to pass the test. Jacquette opposes the criticism of French and takes Turning’s position; therefore, he would find the argument by French to be meaningless. My personal view on the topic is that Robert French was correct to criticise the Turing test due to its vulnerable position to criticism.