共有繁栄インターネット
The Shared Prosperity Internet
自動化とAIの恩恵を社会全体に広く届けるためのネットワークコンピューティングアーキテクチャを提案し、物理的制約から設計原則と技術的柱を導出し、具体的なユースケースと測定可能な成果を定義した論文。
著者: Juan A. Cabrera, Pit Hofmann, Jonas Schulz, Frederic Benken, Hrjehor Mark, Giang T. Nguyen, Holger Boche, Frank H. P. Fitzek
分類: cs.NI, eess.SY
原文アブストラクト
The Shared Prosperity Internet (SPI) is a network-computing architecture that makes the benefits of automation and Artificial Intelligence (AI) broadly accessible to the society. To ground its design, this paper maps the physical constraints of Shannon, Landauer, Turing, and Einstein to three design principles: trustworthiness, sustainability, and technological sovereignty, and maps them into three technical pillars: i) post-Shannon, goal-oriented communication that transmits only what the task requires; ii) anticipatory decision-making ("negative latency") with confidence-bounded pre-action and correction; and iii) beyond-digital computing that selects energy-optimal substrates under deadline and computability constraints. The SPI is grounded in three societal use cases: remote teaching for pupils, remote teaching of robots and cyber-physical systems, and elder care. Furthermore, this paper defines measurable outcomes for an SPI, including latency decomposition, bits per event, energy and CO2 per task, safety and privacy indicators, and robustness.