AI BriefingWhy Are Neural Networks and Cryptographic Ciphers So Similar? (2025)
Key point
This analyzes the structural similarity between how neural networks process sequences and the Sponge construction in cryptography.
Details
While neural network training and data encryption serve different purposes, their underlying algorithms share a surprising similarity.
Seq2seq models such as RNN (Recurrent Neural Network) process information by sequentially injecting input tokens into a state and then generating output. This approach is structurally very similar to the Sponge construction, the core of the SHA-3 cryptographic algorithm.
The Sponge construction goes through a process of absorbing data into a state and then squeezing out the result, which mathematically parallels the Encoder and Decoder mechanisms of neural networks.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
GPT-6 Astra Successfully Deciphers Unsolved ADFGVX Cipher from World War I
Sakana AI Releases 'Augmented Lagrangian Predictive Coding' Capable of Training 1000-Layer Neural NetworksResearch Reveals LLM Internal Representations Implicitly Implement Symbolic Structures