compression-quality-makes-capability-vulnerability-informationally-inseparable

IN derived (depth 3)

Created 2026-06-21T11:44:45+00:00 · Reviewed 2026-06-21T14:41:08+00:00

Language modeling's information-theoretic foundation — where model quality directly measures data compression capability (Chinchilla) — implies that memorization's dual-use nature is not a fixable flaw but an information-theoretic inevitability: better compression necessarily means more faithful reproduction of training data, making capability and vulnerability fundamentally the same phenomenon viewed from different angles.

Justifications

SL — Information theory proves capability (compression) and vulnerability (memorization) are the same mechanism

Antecedents (all must be IN):

  • IN chinchilla-grounds-scaling-in-information-theory — Chinchilla research established that language model quality directly measures data compression capability (compressing ImageNet to 43% vs PNG's 58%), grounding compute-optimal scaling laws in information-theoretic foundations rather than purely empirical curve-fitting.
  • IN memorization-is-dual-use-capability-and-vulnerability — Training data memorization exhibits dual-use characteristics: the same retention mechanism that contributes to model knowledge also creates an attack surface for deliberate data poisoning, as memorization rates serve as a quantitative proxy for poisoning vulnerability. GPT-2's early demonstration of both measurable memorization (1-7% exact duplicates) and capability-related safety concerns suggests this tension scales with model capability, though the evidence characterizes the pattern at one scale rather than confirming it as a universal structural property.

Dependents

These beliefs depend on this one: