https://doi.org/10.25198/2077-7175-2020-2-74

POST-TURING METHODOLOGY: DESTRUCTION OF THE WALL ON THE WAY TO GENERAL ARTIFICIAL INTELLIGENCE

А. R. Efimov

Robotics Laboratory of Sberbank, RAS Institute of Philosophy, National Research

Technology University «MISiS», Moscow, Russia

e-mail: makkawity@gmail.com

Abstract. This article attempts to comprehensively criticize the «Turing test» and develop quality criteria for new tests for assessing artificial intelligence.

It is shown that the prerequisites for reducing personality and human consciousness, which A. Turing used, reflected the level of development of technology of that period. It is shown that communication between a person and a machine can be considered in four ways: verbal interaction in the physical world, non-verbal interaction in the physical world, non-verbal interaction in the virtual world, and verbal virtual interaction. The Turing test, in fact, describes only the last option. Purely verbal communication between subjects, which involves thinking only using symbolic systems, is a «wall» that excludes the possibility of a transition from complex observable phenomena to an abstract image or concept. But in the first decades of the development of computer technology – the Turing test was a simple, understandable example of the ideal of programming, consistent with the level of technical capabilities.

In the 2010s, there have been qualitative changes in programming.

Firstly, the Turing test was formally passed without creating full artificial intelligence. Today, a machine is bet-ter than a person in symbolic systems, but does not have the ability to compare them with the real world. Secondly, new software tools (deep learning neural networks) have appeared that allow one to partially abstract processes (recognize speech, faces, some images), but not formulate new concepts.

This allows us to put forward new requirements for the test evaluation of artificial intelligence: it must support all forms of communication with a person, abstract images and specify concepts, and also have the opportunity to participate in social practices. It seems logical to compare artificial intelligence with the capabilities of children of different ages. At the same time, the anthropomorphism of the robot in itself will not be an indicator of the creation of artificial intelligence.

Keywords: robotics, artificial intelligence, Turing, Turing test, philosophy of artificial intelligence, symbolic systems, verbal interaction, neural networks.

Cite as: Efimov, A. R. (2020) [Post-Turing Methodology: Wall Destruction on the Way to General Artificial Intelligence]. Intellekt. Innovatsii. Investitsii [Intellect. Innovations. Investments]. Vol. 2, pp. 74–80. DOI: 10.25198/2077-7175-2020-2-74.