Russian Science Foundation has supported Cybernetics Department of MEPhI project “The research of biologically based models of cognitive systems”. A.V.Samsonovich, Professor at the Cybernetics Department №22 and supervisor of the project, has told us about its goals and objectives.

– Grant called “The research of biologically based models of cognitive systems” is dedicated to biologically motivated cognitive architectures, and its main idea is creation of some kind of an equivalent of a thinking human brain in a computer. A comprehensive approach is needed there, different scientific areas are required, and our group members represent them.

Our project is aimed at formulating of principles which will allow making an equivalent of a human thought in a computer. We can’t say that our purpose is unique because nowadays the concept of biologically inspired cognitive architectures is very popular. However, our approach is different from other scientists’ ones. It is different from traditional approaches, those which are based on formal logic (machine theorem proving) and those which require detailed biological realism (computational neuroanatomy and neurophysics) as well as the ones which reduce problems to practically implemented patterns like image recognition, optimization, reinforcement learning and others based on classical neural network models, genetic algorithms, Bayesian networks, fuzzy models etc. and their combinations. This “classics”, although in demand in modern practical applications in sphere of future-oriented developments seems to become out-of-date.

Our aim is to create a full-fledged anthropomorphic system which thinks as a human and can be compatible with a human, can work with a human like a partner, like a performer understanding human needs and not as a machine which needs to be programmed or adjusted several times.

There is a need in a machine talking to which a human will understand if it is a robot or a human. So, it should have human abilities, for example, learn to solve problems. It itself will be a solution to problems.

It sounds very ambitious, but it’s not new. The task to formulate and execute in a computer principles which create systems similar to human thought and brain was made in 50-s of the XX century. But it hasn’t been completed up to the next time. Nowadays machines can win in a chess round with a human or in Jeopardy, recognize speech and talk in a human-like voice but they still can’t walk and see as a human or understand them.

For making a human-like intellectual agent capable of living in the society and be useful in it we need to include 3 components – emotional, narrative and cognitive intellect – on condition that basic elements of cognitive cycle have been included (perception, understanding, solution making and their execution) and there are “outlying” solutions (accumulation and usage of general notions about the world, processing of “natural” speech, planning of task-oriented behaviour). In fact, we already have this basis today. We need to add ability of affects recognition and thinking in terms of human emotions, build narratives and use them to reach aims, then we will get a virtual student in a computer which would be helpful for the society (at least virtually).

Below: the Heads of the project “The research of biologically based models of cognitive systems”. A.V.Samsonovich (on the left) and V.V.Klimov (Cybernetics Department)