The idea to replicate in a machine human consciousness,understood as an immediate first-person subjective experience,becomes one of the greatest challenges of our time.Scientific understanding of the nature of phenome...The idea to replicate in a machine human consciousness,understood as an immediate first-person subjective experience,becomes one of the greatest challenges of our time.Scientific understanding of the nature of phenomenal consciousness may require an expansion of the scientific method itself.While this resolution is postponed,the task of reproducing the consciousness phenomenon in a computer should be approached within the framework of functionalism.Proposed so far theories of this sort appear problematic.An approach to implementing humanlike consciousness is proposed here based on BICA(biologically inspired cognitive architectures)combined with LLM(large language models).The concept is illustrated by example of a virtual tutor.It is argued that final solutions can be obtained in a purely neuromorphic form,which will enable their further adaptation,growth,and evolution.We start from discussion of the ontological problem of consciousness and its temporary nonreductive solution,which makes a scientific study possible.The main hypothesis for the proposed study is formulated based on the BICA Challenge in terms of the eBICA cognitive architecture.Only the limited case of socially emotional consciousness is analyzed as example.It is explained how LLM,being not conscious on their own,accessible via ChatGPT,can be connected to eBICA,used to build a conscious prototype and validate its claimed property.Selected example of a virtual tutor further illustrates how this technology will allow the tutor to develop human-level teacher-student relationships that will increase the efficiency of tutoring.Future prospects for the proposed direction of research include a variety of sentient agents to be created for the many practical human needs with the possibility of their further autonomous evolution.展开更多
基金Supported by Russian Science Foundation(22-11-00213).
文摘The idea to replicate in a machine human consciousness,understood as an immediate first-person subjective experience,becomes one of the greatest challenges of our time.Scientific understanding of the nature of phenomenal consciousness may require an expansion of the scientific method itself.While this resolution is postponed,the task of reproducing the consciousness phenomenon in a computer should be approached within the framework of functionalism.Proposed so far theories of this sort appear problematic.An approach to implementing humanlike consciousness is proposed here based on BICA(biologically inspired cognitive architectures)combined with LLM(large language models).The concept is illustrated by example of a virtual tutor.It is argued that final solutions can be obtained in a purely neuromorphic form,which will enable their further adaptation,growth,and evolution.We start from discussion of the ontological problem of consciousness and its temporary nonreductive solution,which makes a scientific study possible.The main hypothesis for the proposed study is formulated based on the BICA Challenge in terms of the eBICA cognitive architecture.Only the limited case of socially emotional consciousness is analyzed as example.It is explained how LLM,being not conscious on their own,accessible via ChatGPT,can be connected to eBICA,used to build a conscious prototype and validate its claimed property.Selected example of a virtual tutor further illustrates how this technology will allow the tutor to develop human-level teacher-student relationships that will increase the efficiency of tutoring.Future prospects for the proposed direction of research include a variety of sentient agents to be created for the many practical human needs with the possibility of their further autonomous evolution.
