The journal “Economic Strategies” published an article by Alexander Raikov, member of the IIS Supervisory board, on the problem of strong artificial intelligence (AI).
The article notes that “modern AI systems can only recognize, predict and answer questions, but they cannot think, understand, explain and pose problems. Failure to provide an explanation generates distrust of the conclusions drawn from such systems. This was the case when expert systems appeared on the market many years ago (then they dealt with this problem), the same limitation is still characteristic of modern neural networks.
A human is able to make correct and at the same time non-causal (unreasonable) decisions, has intuition, can meditate and fall into a trance, he has a mysterious soul. The idea of consciousness is accessible only to man. Apparently, it is these aspects that are the subject for the possible development of modern AI towards a strong AI. To build it, it is necessary to change the paradigm of AI development.
Strong AI can be a hybrid cyber-physical system of systems. In the new paradigm, it is necessary to explicitly include an observer in the work of the AI system, that is, a human who reflexively and cognitively influences the situation. This inclusion makes the formulation of problems inverse, incorrect, and what is more – in cognitive (non-formalized) spaces. Large-scale evolutionary calculations can be used to solve them. Instead of a logical-linguistic or neural network representation of objects and events, it is necessary to delve into the atomic level of their semantic interpretation. Then the semantics of AI models becomes quantum-relativistic. The continual power of such a semantic interpretation in statics is 30-50 orders higher than that in the traditional one. This is not at all the idea of a quantum computer with a superposition of discrete states of quantum particles, but fundamentally different.
When building the paradigm of strong AI, it is necessary to take into account the phenomena of the collective unconscious, energy singularity, quantum nonlocality, wave-particle duality, thermodynamic and relativistic effects, the relationship of low-temperature plasma with the substance of the brain and body, free fluctuations of the vacuum, spontaneous behavior of natural neurons. It will be necessary to compare the experiment with two slits, when a quantum particle interferes with an infinite number of shadow particles, with the semiotic aspects of the hermeneutics of words. The possible interconnection of cosmic strings with quantum superstrings filling the atoms of the instrumental environment of a strong AI should not remain unnoticed. At the same time, it is necessary to take into account the mysteriousness of the unified theory of the field, dark energy and dark matter. These and many other phenomena do not yet have convincing causal explanations. Apparently, the very process of building a strong AI will help find them.
Thus, a certain hybrid strong AI is most likely capable to put the fundamental questions correctly, to find and substantiate the correct answers to them. It includes collective intelligence and new powerful AI tools. To build it, it is necessary to change the paradigm of creating AI systems, delve into the atomic abyss and rise to space heights. At the same time, this requires the development of the institutional foundations of the collective scientific consciousness, tools for communication and virtual cooperation.”
Alexander Raikov is doctor of technical sciences, full state advisor of the Russian Federation of the 3rd class, laureate of the RF Government Prize in the field of science and technology. He is also a leading researcher at the V.A. Trapeznikov Institute of Control Sciences of the Russian Academy of Sciences, head of the department of cognitive technologies of the MSU National Center for Digital Economy, a professor at MIREA – Russian Technological University.
Text: Journal of Economic Strategies (in Russian)