This page may contain stale information. Last updated: 2026-06-07
Definition
Turing universality means a computational system can simulate any Turing machine and thus compute any computable function, given sufficient resources. Established for synchronous recurrent neural networks in the 1990s (Siegelmann & Sontag).
Key Points
- 2026-06-05: Extended to asynchronous-neural-networks — global clocks not required for universality (2026-06-05-async-neural-networks-nature)
- Prior proofs: spiking neural P systems, transformers with hard attention, finite-precision transformers with chain-of-thought
- Practical significance: full computational power may coexist with energy-efficient async updates
- Distinction: theoretical universality ≠ practical training/deployment efficiency
Related
- asynchronous-neural-networks
- energy-efficiency-in-ai
- neural-network-efficiency
- transformer-architecture