Why AI systems don't learn – On autonomous learning from cognitive science
8 hours ago
- #Autonomous Learning
- #Cognitive Science
- #Artificial Intelligence
- Current AI models have limitations in achieving autonomous learning.
- Proposed learning architecture integrates observation (System A) and active behavior (System B).
- Meta-control signals (System M) enable flexible switching between learning modes.
- Framework inspired by human and animal cognition for real-world adaptability.
- Discusses evolutionary and developmental timescales in learning adaptation.