Lifelong Forgetting: A Critical Ingredient of Lifelong Learning
← Back to Publications
Lifelong Forgetting: A Critical Ingredient of Lifelong Learning, and Its Implementation in the OpenCog Integrative AI Framework
Authors:
Year: 2011
Venue: AAAI Workshop (Lifelong Learning) (2011)
Links: paper (PDF)
Summary
Argues that forgetting — framed technically as “memory access-speed deprioritization” — is among the most central ingredients distinguishing lifelong learning from shorter-term learning, since faster-access memory remains considerably more expensive than slower-access storage. It reviews the shades of meaning of forgetting and discusses the forgetting mechanisms used in the OpenCog integrative cognitive architecture (deciding what to keep in fast memory vs push to slower stores, with outright deletion as one extreme).
Relevance to Hyperon
An attention-economy source for the ECAN Deep Dive: forgetting in OpenCog is governed by the importance/attention-allocation mechanism (low-importance atoms are deprioritized or forgotten), directly relevant to ECAN's resource-bounded importance dynamics and memory management.