Will THIS replace LLMs? O(n²) problem

TLDR; Transformers has a quadratic complexity problem: O(n²). Building more data centers around the silicon microchip to solve an algorithmic issue is a ultimately a brute-force dead-end. Reservoir computing not only escapes O(n²) entirely, shifting from LLM’s static word predictor toward a native physical, dynamic, chaotic predictor. Add the speed of photonic computing and the efficiency of neuromorphic computing, we have an emerging winner: a new domain of Intelligent Embodied AI.