Overview
Neuromorphic metacognition explores an independent computational system that observes AI behavior rather than relying on the AI system to evaluate itself.
PROJECTS
Neuromorphic Metacognition
An independent brain-inspired computational architecture for observing artificial intelligence and generating graded indications of concern.
Observation
The architecture receives behavioral evidence from an AI system and represents changing patterns across a separate computational substrate.
Recognition
Formal and learned representations can be used to identify patterns associated with uncertainty, inconsistency, manipulation, hallucination, and other states.
Graded Concern
Rather than issuing only binary judgments, the system can communicate degrees of concern to a human user.
Human Protection
The central purpose is independent oversight that provides humans with additional information when deciding whether and how much to rely on AI output.