model-collapse-recursive-crisis-amplifier
IN derived (depth 7)
Created 2026-06-21T14:08:47+00:00 · Reviewed 2026-06-21T15:37:01+00:00
Model collapse from synthetic data creates a recursive amplifier within ML's compounding reliability crisis — as capable models generate training data for next-generation models, reliability degradation is inherited and compounded across model generations, meaning capability scaling now directly poisons the data substrate on which future capability depends, adding a temporal feedback dimension to the crisis.
Justifications
SL — Synthetic data creates intergenerational reliability degradation within the compounding crisis
Antecedents (all must be IN):
- IN ml-model-collapse-synthetic-data — Model collapse is the degradation that occurs when models train on uncurated synthetic data or outputs of prior model versions, also called 'model autophagy disorder (MAD)'
- IN reliability-crisis-compounds-with-capability — ML's reliability challenges appear structurally related to its capability gains — the pragmatic, hardware-driven scaling that selects for architectures achieving strong performance may also contribute to characteristic failure modes like adversarial fragility, while standard evaluation methods and existing paradigms fail to detect or eliminate the resulting bias and robustness vulnerabilities. This suggests a persistent gap between demonstrated capability and deployment trustworthiness that current methodologies do not adequately address, though the link between capability-driving factors and fragility-introducing factors reflects correlation and plausible connection rather than a fully established causal mechanism.
Dependents
These beliefs depend on this one:
- IN compute-revolution-contaminates-own-data-supply — The 300,000x compute increase that drove ML's capability revolution simultaneously creates the conditions for model collapse — massive compute enables training on internet-scale data that produces capable models, but those capable models flood the internet with synthetic content, contaminating the very data ecosystem that enabled the scaling in the first place.
- IN model-collapse-adds-data-scale-to-fractal-crisis — Model collapse from synthetic data may represent a fourth self-reproducing scale in ML's fractal crisis pattern — a data generation layer where synthetic training data recursively degrades model quality. This would parallel the three scales already identified (model-level via GANs, paradigm-level via generative churn, field-level via pragmatism dynamics), though whether data-level degradation reproduces the identical structural pattern or constitutes a distinct recursive mechanism is not yet established by the evidence.