Watchmaker Functions and Meta Specification of Open-Ended Learning Systems
Keywords: Open-ended learning systems
Abstract: Open-ended learning systems aim to foster the continuous evolution of increasingly capable agents through the dynamic generation of novel challenges. The efficacy of these systems is fundamentally influenced by two critical factors: the design of the underlying system, which delineates the space of possibilities, and the open-ended algorithms that drive ongoing progress within this space. Current approaches to system design rely on explicit specification, where state spaces and evolution functions are fully defined at design time, often leading to prohibitive design complexity as systems scale. To address this challenge, we propose an alternative design principle termed *meta specification*. This approach defines systems implicitly through constraints, utilizing *watchmaker functions*—generalized stochastic evolution functions—coupled with verification routines to perform system evolution. Meta specification principles have the potential to significantly expand the space of possibilities while reducing design complexity, thereby enhancing the potential for open-ended learning. We demonstrate the viability of this principle through an illustrative implementation that co-evolves robot morphologies and robotic tasks, showcasing its capacity for emergent novelty and highlighting the shift in focus towards verification in system design.
Primary Area: transfer learning, meta learning, and lifelong learning
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 13036
Loading