Chronoseed: A Generative AI Time Capsule
This paper assembles a design space and defines the parameters for a generative time capsule—one that preserves not only an archive of human knowledge but also the cognitive function of human thought within a neural network. Such a time capsule must persist across vast timescales, including scenarios of civilizational collapse, while providing mechanisms for decoding its contents and executing its program, even in radically altered technological and epistemic contexts. We introduce Chronoseed as a speculative thought experiment, exploring the possibility of an AI time capsule that preserves both stored knowledge and the ability to process it. To propose its form and evaluate its feasibility, we construct a three-dimensional design space—durability, accessibility, and completeness—as a framework for systematically examining how synthetic intelligence might reshape cultural preservation. By engaging with historical time capsules, long-term nuclear waste warning systems, and speculative contributions from science fiction writers, the paper considers the broader challenge of transmitting knowledge and logics of thinking across deep time. It positions Chronoseed’s generative function as an alternative to static repositories, proposing a model in which knowledge remains interactive, adaptive, and capable of being reconstructed by future intelligences. Using the assembled design space and its parameters, the article explores possible physical forms of Chronoseed, balancing preservation with the ability of future discoverers to activate its generative potential. It ultimately identifies DNA embedding as the most viable option for long-term durability and recoverability.
Authors
- Jackie Kay (ORCID: https://orcid.org/0000-0001-9593-695X)
- Winnie Street
- Sonia Bernaciak
Institutions
- University of London (GB)
- Royal College of Art (GB)
- Google (United Kingdom) (GB)
- University College London (GB)
Publication Details
- Published
- 2025-05-04
- DOI
- https://doi.org/10.1162/anti.5czo
- Primary Topic
- Semantic Web and Ontologies
- Type
- article
- Field-Weighted Citation Impact
- 0.00