Every day, AI systems based on neural network algorithms are solving a growing variety of tasks, and despite their complexity, the quality of solutions continues to improve. The article discusses why AI systems capable of solving a wide range of problems still do not meet the expectations of AGI. Several definitions of AGI are compared, and five significant shifts in AGI design perspectives are presented, the adoption of which could bring AI systems closer to aligning with AGI definitions.

artificial intelligence (AI), neural network algorithms, strong (general) AI (AGI), approximation, knowledge, decomposition, complex environment, dynamic model