The Unstable Artifact
Traditional preservation relies on the "fixity" of the object. A book does not change its text when you close it. A Generative System does. Examples include:
- Procedural Video Games: The world of No Man's Sky (2016) is generated by a mathematical formula. If the developers patch the formula, the universe physically changes. The "old" universe is not just deleted; it is mathematically impossible to revisit.
- Algorithmic Feeds: Your TikTok "For You" page is a unique, ephemeral cultural artifact generated for you at a specific micro-moment. It exists for seconds, then vanishes.
Preserving the Probability
How do we archive a system that changes every time you look at it? We replace "Object Preservation" with "Logic Preservation."
- Snapshotting: Recording hours of gameplay or output to capture a "representative sample" of the system's behavior.
- Source Code Audit: Preserving the algorithm itself (the seed) rather than the output (the tree). If we have the code and the seed value, we can theoretically regenerate the universe.
Field Notes
The Minecraft "Seed": Minecraft players share "world seeds"—alphanumeric strings that tell the game's terrain generator exactly how to build a specific world. This is a form of compressed geography. Passing a seed to a friend is like handing them the DNA of a continent.
Generative Archaeology: Some researchers use generative AI to "fill in the blanks" of physical ruins, training models on existing pottery shards to predict the shape of missing ones. This creates a "Probabilistic Past"—a history that is statistically likely but factually unproven.
Ephemera
A major challenge with Generative AI (LLMs like GPT) is "Model Collapse." As models are trained on the internet, they begin to ingest AI-generated content (their own vomit). This causes the system to degenerate into nonsense. Archaeobytologists track these "inbred datasets" as markers of the "Synthetic Era."