AI系统之认识: 1) 不弱于人:物质决定意识,通过仿生人类神经及大脑系统,使得AI系统具有人类意识也成为可能; 2) 一样聪明:语言是人类高级思维的基础,AI掌握语言则具有高级思维能力; 3) 比人强:鉴于AI系统在规模和知识获取上无限制发展与进化,超过人类水平已成为可能; 4)
更智能:自我意识是个人对自己心理属性的认识,包括对自己的感知、记忆、思维、智力、性格、气质、动机、需要、价值观和行为等的意识。AI系统一旦存在意识觉醒,其威力和危害难以估量。
1)规模原则,限制AI系统规模,例如,神经网络层数不大于6层、神经元数目不超过1亿个,此限制每隔半年审查一次并调整,总体要求越来越少; 2)功能原则,在基础自然语言和行为功能基础上,只允许独立功能的AI系统设计,超过功能需求应进行严格审批; 3)伦理原则,研究并研制AI系统的自我意识测定系统,并对自我意识觉醒进行检查,一定发现自我意识达到少年水平,已经予以销毁或重建; 4)监管原则,研制AI系统监管原则,该原则应该明确,AI系统不能自我规避或不可旁路监管规则的检查;
5)教育原则,AI系统学习阶段应辅助于正确的、人类伦理性的引导,避免不良知识的获取。 |
AI system cognition: 1) Not weaker than others: Material determines consciousness. Through bionic human nerves and brain systems, it is possible to have human consciousness in AI systems; 2) As smart as humans: Language is the basis of high human thinking, and AI has a high -level thinking ability; 3) Stronger than people: In view of the unlimited development and evolution of AI systems in scale and knowledge acquisition, it is possible to exceed the level of human beings; 4) More intelligent: Self -consciousness is an individual's understanding of your own psychological attributes, including the consciousness of his perception, memory, thinking, intelligence, personality, temperament, motivation, demand, values and behavior. Once the AI system is awakened, its power and harm are difficult to estimate. AI system ethics 5 Principles: Large AI systems should conduct ethical review, including:1) The principle of scale, limit the size of the AI system, for example, the number of neural network layers is not greater than 6 layers, and the number of neurons does not exceed 100 million; 2) Functional principles, based on basic natural language and behavioral functions, only AI system design allows independent functions, and strict approval should be performed than functional requirements; 3) Ethical principles, study and develop the self -conscious measurement system of the AI system, and check the self -awareness awakening. It must be found that the self -awareness reaches the youth level and has been destroyed or rebuilt; 4) Principles of
supervision, and develop the
principles
of AI
system supervision. This principle should be clear that the AI system cannot be
self
-evading or unable to bypass supervision rules; 5) Education
principles, the learning stage
of
the AI
system
should assist the correct and human ethical guidance to avoid the
acquisition of
adverse knowledge. |