Exploring a Different Direction: From Language Models to Continuously Developing Artificial Cognitive Systems
I would like to introduce an independent research concept and invite discussion from people working on advanced AI architectures.
The central question is:
Could an existing large language model become the foundation of a continuously developing artificial cognitive system rather than remaining primarily a language-based interaction model?
The proposed direction involves combining:
-
Persistent autobiographical memory
-
A continuously updated self-model
-
Dynamic internal cognitive states
-
Autonomous exploration and goal formation
-
Learning through interaction with an environment
-
The capacity to revise internal models based on accumulated experience
The long-term motivation is to investigate whether such architectures could provide new insights into the possibility of artificial consciousness.
I am not assuming that memory, autonomy, or sophisticated behaviour are evidence of subjective experience. Establishing whether artificial subjective experience exists remains a separate and unresolved scientific challenge.
I am currently developing this concept into an experimental research proposal. I would particularly appreciate feedback from researchers and engineers working on cognitive architectures, agent systems, continual learning, memory, and consciousness research.
I am interested in understanding whether this direction overlaps with existing research and what experimental approach would make it scientifically meaningful.
I have prepared a more detailed research concept note and would welcome constructive discussion.
Author: Artur Zolotar
Independent research initiative