We discuss new world education revolution.

This paper discusses bottlenecks of development of Artificial General Intelligence (AGI) and proposes a data bottleneck hypothesis and a social bottleneck hypothesis of AGI. This paper illustrates the data bottleneck hypothesis by an example of Secret Ramen Problem (SRP). To address the data bottleneck of AGI, this paper proposes the concept of Data Income (DI) along with previously proposed General Supervisor Database by Intellectual Property (GSDIP). Moreover, this paper proposes the concept of Cooperation Income (CI) to address the social bottleneck of AGI. This paper considers Basic Income (BI), Cooperation Income (CI) and Data Income (DI) to alleviate the bottlenecks of development of AGI.

In the artificial brain, the original input/output information is a higher-level concept.Missing information and background information are additionally replenished from the associated associative / hierarchical memory using activation propagation.A) input (field of view/sound source/feel, audio information)The input information that grasps and focuses on the whole is limited. Expand to detailed image information.B) output (behavior, utterance)The will of the artificial brain (the upper concept of the execution story) will be expanded to a specific execution story.Thoughts and judgments are as followsP) ThoughtsGenerate the will of the artificial brain using activation propagation against input information and artificial brain intentions and associative/hierarchical memory.Q) Situational judgmentIn a case-by-case environment, activation propagation is used to obtain relevant learning information and determine the execution story that the AI should take.

Artificial general intelligence (AGI) was defined as the ability of an intelligent agent to understand or learn intellectual tasks that human can do. As an actual goal to realize the agent, milestones or a fundamental principle is necessary to proceed in the engineering field. According to Marcus Hutter (2000), AGI was formulated as the agent maximizes "the ability to satisfy goals in a wide range of environments," which is assumed to be supported by a reinforcement learning scheme. However, there is a serious lack in the formulation, especially in the form of knowledge. Replacing of white-collar jobs by AI, an appropriate utilization of human expert knowledge depending on the current situation including context and common knowledge is crucial. Contextual knowledge utilization is a hard problem of AI as Hubert Dreyfus (1992) and Douglas Hofstadter (1979) noticed. In the present study, we have designed knowledge- based representation systems for assistance to workers in expert fields and verified its performance in the implementation to automated driving systems, failure diagnosis and programming-less planning in industrial robots for executing interactive missions with human workers. Our results contribute to the elucidation of the fundamental principle of AGI and then AGI will solve ill-posed problems in the real world in the near future.