摘要
专用人工智能技术不断推进,但类人思维的人工智能技术仍然亟待突破,未来人机共同参与的社会治理亦面对重重挑战。这些问题都要求我们必须更加深入探寻人机本质,在“人更像机器”还是“机器更像人”之间做出抉择。我们主张后者才是对人机未来更有益的发展方向。要实现这一目标,就需要探讨人类意识能否传递以及如何传递给机器。对人类而言,意识单元(我们称之为“认知坎陷”)的产生与身体相关,如果意识脱离了人的身体上传给机器,就需要对意识分级并讨论其可迁移的程度。物理数学规律具有绝对可迁移性,意识世界的产物具备相对可迁移性。人能够将抽象概念通过附着与隧通进行具象表达并优化,在人际、代际间传播并达成共识,即形成了具备可迁移性的认知坎陷,而“自我”就是其中最原初的意识单元。通过认知坎陷工程化的方式,有可能让机器形成“自我”原型,习得隧通与迁移,实现机器类人思维的突破,即使如此,人的意识也无法完整地上传给机器。人机共融将有可能在第三代互联网或元宇宙中率先进行尝试并有望实现。
Artificial Intelligence technology of specific disciplines continues to surge,while the development of AI technology with human-like thinking still calls for breakthroughs.Confronting with the governance challenges of the human-computer-coexistence society in the future,it is imperative to explore the human nature compared to machines,and to make a choice between making human beings more machine-like or making machines more human-like.Of which the latter was preferred.It requires us to understand whether and how human consciousness is uploaded to machines.A man can transform and optimize abstract concepts into concrete expressions,which is identified as cognitive attractors.They are transferable among people and between generations by reaching consensuses.Human consciousness,which originates from human bodies,can be scaled based on its transferability.For example,the physical laws and mathematical theorems are absolutely transferable,and other cognitive attractors are relatively transferable. “Self”is the most important and primitive cognitive attractor and can be concreted by bandwagoning and tunneling to other cognitive attractors,which provides an engineering possibility for machines to form their prototypes of “Self”,and to acquire human consciousness.However,human consciousness can not be completely uploaded to machines even though the breakthrough of human-like thinking machines is achieved.Nevertheless,human-computer-coexistence will be firstly attempted and actualized in Web 3.0 or Metaverse.
作者
蔡恒进
蔡天琪
Cai Hengjin;Cai Tianqi
出处
《社会科学战线》
CSSCI
北大核心
2022年第3期154-161,282,共9页
Social Science Front