A.Yu. Alekseev - Ph.D. (Philos.), Associate Professor, Moscow State Institute of Culture. E-mail: aa65@list.ru

Patricia and Paul Churchlands positively responded to the fundamental question of philosophy of artificial intelligence (AI), «Can machines think?» [1]. But demanded to improve a machine that is able to simulate intelligent behavior: such a machine should simulate human brain activity. Some researchers thought that the «Turing test to the brain» is extremely effective. For example, it allows to argue plausibly philosophical zombie (O. Flanagan and T. Polger). In our view, the Churchlands test developed hastily and not completed. Firstly, P. and P. Cherchlendy not attentive to the argument H. Dreyfus, fundamentally deny the possibility of AI. «The development of AI goes hand in hand with the ever-increasing processing power, memory capacity and complexity of the knowledge base» - they believe [1, P.33]. But do not overlook the fact that one of the main problems, identified Dreyfus, is the inability of a computer to form a distinct «knowledge» on the basis of background, implicit «knowledge». For example, the computer is not able to handle metaphors. That cannot be overcome from quantitative achievements of computer technology. Second, they believe that the criticism of the original Turing test from Neurocomputing test contributes comparable with the Searle Test («The Chinese Room»). Turing machine is a symbolic discrete machine. That machine is unable to represent the causal relationship between mental phenomenon and continuous neurophysiological functions. Requires a Turing machine to replace a neurocomputer. However, such a hypothetical statement requires verification. John Searle appeals to the intuition of phenomenal consciousness, to the first person necessary judgments. Therefore, the epistemological significance of his work is much higher. Third, Churchlands believe that Church is the founder of the idea of intelligent machines, which plays in the Turing game of imitation. Intelligent functions represented by efficiently computable functions, as any player response to a question the judge can be obtained within a reasonable time and with limited resources. These functions, in turn, reduced to the recursive functions, and they are based on primitive recursive functions. Sequential Turing machine provides engineering formalization ideas A. Church. However, in imitation of the intellect is played not the Turing machine (1936), but the Universal digital computing machine (1950). It includes some elements of probabilistic parallel computer. For example, a Rulebook used which consists from the set pages (tapes) superposed [2, P.44-46]. Moreover, to describe the work of a Child-Machine A. Turing uses neural network! [2, P.231-237]. Fourth, do not understand the specifics of the Churchland\'s «Neurocomputing». It turns out that the neural system - a parallel, failover, persistent, high-speed machines [1, P.35-36]. The main thing is that the way they function is not based on the manipulation of symbols, but to vector conversion of continuous values. However, where there Neurocomputing - Such a machine independent parallel Churchland represented by Turing machines, which are controlled by a multi-dimensional automaton table in accordance with the theory of formal neurons. Between the tapes of these machines lack of internal communication. There are only external relations. They agreed linguistic complexes recorded on tape parallel. It turns symbolic paradigm AI complicated meta-language of the theory of neural networks. Connectionism is missing! Thus, it seems that Churchlands inattentive and not fully read the article of A. Turing. Connectionist paradigm is not visible. Neurocomputing Turing test worked surface and requires clarification. For his development of promising use of the Korsakov-Turing machine [3]. Connectionist «computability» organized on subsymbolic level through internal connections between the Turing machine tapes. This work was financially supported by RHSF grants № 15-03-00519а «Postnonclassic paradigm of Аrtificial Intelligence».

 

1. Churchland Paul M., Churchland P.S. Could a Machine Think - // Nature Publishing Group. V. 262. № 1. P. 32-37.
2. Alekseev A.JU. Kompleksnyjj test Tjuringa: filosofsko-metodologicheskie i sociokulturnye aspekty. M.: IInteLL. 2013. 300 s.
3. Alekseev A.JU. Protonejjrokompjuter Korsakova // Nejjrokompjutery: razrabotka i primenenie. 2013. № 7. S. 6-17.