rnn-turing-completeness-purest-case-of-theoretical-irrelevance
IN derived (depth 5)
Created 2026-06-21T14:08:47+00:00 · Reviewed 2026-06-21T15:37:01+00:00
RNNs' proven Turing-completeness (Siegelmann & Sontag 1994) — the strongest computational-theoretic result for any neural architecture family — is the purest demonstration that theoretical computational power is irrelevant to paradigm survival, as Transformers displaced RNNs solely through superior hardware utilization despite possessing strictly less formal computational power.
Justifications
SL — The strongest theoretical result in neural computation failed to prevent paradigm displacement
Antecedents (all must be IN):
- IN rnn-turing-completeness-siegelmann-sontag — Siegelmann & Sontag (1994) proved that recurrent neural networks with rational weights are Turing-complete, and with real-valued weights they are super-Turing.
- IN paradigm-survival-determined-by-scalability-not-theory — Mathematical completeness and theoretical elegance are neither necessary nor sufficient for paradigm survival in ML — GANs had the most complete analytical characterization yet were eclipsed by diffusion models, SVMs had convex guarantees yet were outscaled by neural networks, while theoretically less grounded approaches that scaled with hardware thrived.
Dependents
These beliefs depend on this one:
- IN rnn-turing-completeness-and-svm-bayes-optimality-jointly-irrelevant — ML's two strongest mathematical results — RNNs' proven Turing-completeness (the strongest computational-theoretic result for any architecture family, made irrelevant by Transformer displacement) and SVMs' Bayes-optimal classification (the strongest statistical-theoretic result, made inaccessible by scaling barriers) — are jointly irrelevant to the field's trajectory, establishing that mathematical optimality at both the computational and statistical levels is independently orthogonal to paradigm survival.