The mathematician Alan Turing, famous for his World War II code-breaking exploits, is also well known for offering a procedure for determining whether true intelligence can be judged to inhabit the circuits of a computer. Known as the Turing Test, it is a simple elaboration on the Imitation Game, in which a player is required to guess the correct identity of a particular human subject given two individuals, each posing as this person. The subject and the imposter are both hidden from the player’s view and the player has an allotted time frame to ask the subject, as well as the imposter, as many probing questions as needed before attempting to guess who the real subject is. Turing proposes that if we replace the human imposter with a digital computer imposter, while imposing the same game rules, then the computer can be considered truly intelligent if it can fool the player into thinking it is actually a human. If we cannot tell the difference between human and machine through conversation, then the machine must be granted the power of intelligent thinking.
Philosopher John Searle has offered a counterargument to the Turing Test. Just because a player in a game can be fooled into thinking a machine can think does not mean that machines truly possess intelligence, or at least an intelligence that could be considered an equivalent to human understanding. As an analogy to the problem, Searle describes an isolated room in which an English speaker, with absolutely no comprehension of Chinese, is equipped with a complex system of reference materials and explicit rules about the Chinese language. Outsiders can feed written Chinese questions into the room, and the person in the room is able to employ the system of references and rules to assemble the proper Chinese responses for output. The person has no clue what the Chinese characters, which are input and output, actually discuss, but to the Chinese speakers outside the room this is not apparent at all. Even though the Chinese participants are fooled, the person in the Chinese Room is not truly able to understand Chinese the way she truly understands English. Thus the Turing Test is insufficient for proving Artificial Intelligence, at least in a way that adequately represents human understanding.
Searle suggests that the only way that we will be able to truly recreate the uniquely human experience of understanding is by synthesizing the entire nervous system, instead of trying to impersonate people with digital computer models. The computer metaphor for the brain, in which the mind is depicted as software, does not literally describe the way that human consciousness emerges from the physical substrates of the nervous system. The functions of intelligence and understanding are possible through the collaborative interplay between many parts of the brain, from the un-conducted orchestration of dispersed and specialized regions, to the complex firing of neuronal networks and their plasticity. Just to process language, humans need at least two distinct parts of the brain, Wernicke’s Area and Broca’s Area, which are used for semantic recall and grammatical understanding respectively. Connectionism proposes that in order to artificially reproduce human understanding, then the brains parallel processing capabilities, and the intricate interaction of neuronal networks must be replicated. To this end, building computer programs that are designed to learn through carrot and stick rules has been investigated. For this alternative outlook, advances in computer science and technology bring us closer and closer to the potential of creating a software program that will pass the Turing Test, Searle’s objection to its explanatory power notwithstanding.
Searle’s Chinese Room thought experiment is basically an articulation of the Argument from Consciousness, which Turing (52) already addressed and credited to Professor Jefferson. Turing contends that this argument is solipsist, because it entails that communication is insufficient to determine if another human is conscious, and requires the impossible standard of directly experiencing the mind of another entity for this verification. Turing is correct that the best we can do is device a test, whether for Artificial Intelligence or Artificial Consciousness, but to accept true machine consciousness, in the same way that a human is conscious, the Turing Test is just not enough. Professor Jefferson suggests that a machine should be able to “write a sonnet or compose a concerto” (Turing 52) in order to be considered conscious, so that we would know that it had felt the emotions and drives that are required for such artistic feats. I would like to meet somewhere in the middle for testing synthetic consciousness, especially since many normally conscious people are likely to fail a test that required them to compose a concerto.
Music, like language, is processed by many disparate regions of the brain working in tandem. Rhythm, tempo, melody, harmony, volume, and lyrics all require attention, let alone the memories, emotions, and physical sensations that flood our consciousness when we hear a song we resonate with. Almost every region of the brain and neural subsystem is utilized for the appreciation of music (Levitin 84). The transcendental experience of getting absorbed into a piece of music defines a subjectivity that is difficult to imagine a machine copying, and if it could this would be the ultimate demonstration of Artificial Consciousness. Therefore, I propose the Endicott Test. Once a computer program has passed the still important and needed Turing test, it and the human it is pretending to be, must then also listen to three songs of my choosing, two songs they have listened to previously, and one that they have never heard before. After a thirty minute discussion about the songs, regarding such topics as genre, style, lyrics, beat, guitar, vocals, catchiness, performance, emotional impact, associated memories, and general appreciation, I think I could easily spot the human. If a computer can fool me with this additional test, then I think we have good evidence for Artificial Consciousness.
Searle’s intuition that Artificial Consciousness needs more than a digital computer is certainly correct. Although, his solution of modeling the nervous system may be necessary, but not sufficient, in order to pass the Endicott Test, even if this method passed the Turing Test, if the elements required for consciousness are not in the neurons themselves. Consider a promising explanation for phenomenal human consciousness proposed by Johnjoe McFadden called the Conscious Electromagnetic Information (CEMI) Field Theory:
Digital information within neurons is pooled and integrated to form an electromagnetic information field in the brain. Consciousness is the component of the brain’s electromagnetic information field that is transmitted to motor neurons and is thereby capable of communicating its state to the outside world. (McFadden 17)
Evidence of the evolutionary benefits of a CEMI field for learning and intelligence, as well as a potential path for McFadden to test his theory through the development of Artificial Consciousness, can be found in an experiment by the School of Cognitive & Computing Sciences at Sussex University. The team used a field-programmable gate array (FPGA), a silicon chip with cell arrays and programmable switches. Using software designed to evolve the system, in order to find the most efficient way to distinguish between two tones, the network was eventually able to perform this task with only 32 of the possible 100 FPGA cells. This optimal point was reached after 5,000 generations. It turns out that some of the cells necessary to performing this task were not physically wired to the others, yet did influence the optimality of the system. This indicates that the system had evolved to use electromagnetic field effects, a by-product of the programmed current moving in the wires, to achieve the most advantageous configuration (McFadden 53-54). McFadden argues that the electromagnetic field effects inside human brains are the actual substrate of human consciousness and subjective awareness, for which experiments that test for consciousness in electromagnetic field effects could provide a viable research agenda. Could investigations into CEMI theory lead to machines that can pass the Endicott Test?
In the meantime, it has been absolutely established that robots can play the drums:
Jared Roy Endicott
Subscribe in a reader
Levitin, Daniel J.This is Your Brain on Music. New York: Dutton, 2006. Print.
McFadden, Johnjoe. “The Conscious Electromagnetic Information (CEMI) Field Theory.”. Journal of Consciousness Studies , 9 (8), pp.45-60. 2002. Web.
Searle, John R.. “Minds, Brains, and Programs.” The Philosophy of Artificial Intelligence . Ed. Margaret A. Boden. New York: Oxford University Press, 1990. 67-88. Print.
Turing, Alan M.. “Computing Machinery and Intelligence.” The Philosophy of Artificial Intelligence . Ed. Margaret A. Boden. New York: Oxford University Press, 1990. 40-66. Print.