Why AI systems don't learn and what to do about it: Lessons on autonomous learning from cognitive science
🤖 AI Summary
Current AI systems lack the capacity for sustained autonomous learning in dynamic environments. To address this limitation, this work proposes a cognitive science–inspired dual-system learning architecture: System A supports observational learning, while System B drives active behavioral learning, with a meta-control module (System M) dynamically orchestrating the two. This framework formally operationalizes biological adaptation mechanisms—spanning both evolutionary and developmental timescales—as a computable model, integrating cognitive modeling, meta-control theory, and reinforcement learning to achieve biologically plausible continual learning. By unifying these perspectives, the study establishes a theoretical foundation and a practical pathway toward next-generation autonomous AI systems capable of open-ended adaptation in complex, changing environments.
Application Category
📝 Abstract
We critically examine the limitations of current AI models in achieving autonomous learning and propose a learning architecture inspired by human and animal cognition. The proposed framework integrates learning from observation (System A) and learning from active behavior (System B) while flexibly switching between these learning modes as a function of internally generated meta-control signals (System M). We discuss how this could be built by taking inspiration on how organisms adapt to real-world, dynamic environments across evolutionary and developmental timescales.
Problem
Research questions and friction points this paper is trying to address.
autonomous learning
AI systems
cognitive science
learning from observation
active behavior
Innovation
Methods, ideas, or system contributions that make the work stand out.
autonomous learning
meta-control
cognitive architecture
learning from observation
active learning
🔎 Similar Papers
No similar papers found.
