On the task approach in artificial intelligence

The authors discuss the problem of the integration approach to artificial intelligence, analyzing the content and positive aspects of the integration agent approach. It is noted that this approach implicitly follows the task approach. The paper gives answers to the questions that make up the essence of the task approach - where do the tasks come from, what is the task, what should be considered a solution to the problem. It also discusses the classification of intellectual problems into direct, inverse, and hybrid. It is noted that modern artificial intelligence focuses mainly on solving direct and inverse problems, leaving a huge and important class of hybrid problems
outside its scope of attention. The paper describes the theoretical model approach to solving the whole variety of intellectual problems, called semantic modeling. It analyzes the advantages of the proposed conception, including the possibility of a flexible combination when solving hybrid problems of tools already created in artificial intelligence. It also discusses the problem of creating a “strong” / “general” artificial intelligence (AGI) in the framework of the task approach.

Keywords: artificial intelligence, expert systems, machine learning, neural networks, deep learning, agentbased
approach, task approach, direct, inverse and hybrid task, “strong” artificial intelligence (AGI), computability, axiomatic and model-theoretic approach, semantic modeling.