Slavin Boris B
Financial University Under the Government of Russian Federation, Moscow, Russia.
Heliyon. 2023 Mar 10;9(3):e14443. doi: 10.1016/j.heliyon.2023.e14443. eCollection 2023 Mar.
This study presents an architectural approach for building a conceptual model of artificial general intelligence (AGI). The architectural approach is generally used to model information systems (IS) of enterprises and can be also used as part of a system-wide approach to describe other complex open systems. The paper suggests three layers and five levels of the AGI model. Two levels (entropy and process) are at the technological layer of AI functioning, two more levels (social and linguistic ones) are at the relationship layer responsible for the behavior of AI, and, finally, the uppermost level (actualization) supposes general intelligence. All the components of each upper layer are connected to the components of the lower layers forming the AGI model. The feature of the social layer is determined by the requirements to the subjectivity of the intellect, its ability to make decisions independently and be responsible for them. The task of the upper layer is self-identification of AGI and understanding its place. The hypothesis has been put forward that the limitation of the life cycle is an important condition for the actualization of intelligence.
本研究提出了一种构建通用人工智能(AGI)概念模型的架构方法。该架构方法通常用于对企业信息系统(IS)进行建模,也可作为系统范围方法的一部分,用于描述其他复杂的开放系统。本文提出了AGI模型的三层和五级结构。其中两层(熵层和过程层)位于人工智能运行的技术层面,另外两层(社会层和语言层)位于负责人工智能行为的关系层面,最后,最上层(实现层)假定为通用智能。每个上层的所有组件都与下层的组件相连,从而形成AGI模型。社会层的特征由对智能主观性的要求、其独立决策并为之负责的能力所决定。上层的任务是AGI的自我识别及其自身位置的理解。有人提出假设,即生命周期的限制是智能实现的重要条件。