arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2603.21795cs.CY
收藏

乘坐列车:非洲在现代人工智能的十字路口

Take the Train: Africa at the Crossroad of Modern AI

Carnegie Mellon University Africa(卡内基梅隆大学非洲分校);Carnegie Mellon University(卡内基梅隆大学);University of Washington(华盛顿大学)

Cédric Manouan, Miquilina Anagbah, N'guessan Yves-Roland Douha, João Barros

AI总结:

本文探讨非洲在现代AI发展中的基础设施和政策瓶颈,提出通过构建稳健的数字基础来实现可持续AI采用,并介绍Africa AI Compute Tracker工具以促进透明度和访问性。

摘要:

非洲参与现代AI发展受到严重基础设施和政策缺口的限制。关键障碍包括有限的高性能计算(HPC)访问、由于支付系统不匹配导致的云访问受限、汇率波动以及严格的数据主权法律,这些法律阻碍了非洲联盟(AU)成员国之间的区域合作。尽管像Cassava AI的AI工厂网络等倡议显示了非洲对AI的兴趣,但这些项目仍较为局部化,而大陆层面的采用仍需更好的协调。本文基于官方声明,提供定性和定量证据,表明可持续的AI采用需要通过平衡的计算、数据和能源访问来建立稳健的数字基础。我们称之为“权利使能器”,认为它们是当前全球AI竞赛中成功的关键组成部分。我们还介绍了Africa AI Compute Tracker(ACT),一个交互式地图,用于监控整个大陆AI-ready HPC系统的可用性。该工具代表了首个开源努力,旨在整合非洲不断演变的HPC景观数据,并鼓励本地AI利益相关者提高透明度,同时促进更广泛的AI开发者访问。本文的工作强调了缩小AI鸿沟并让非洲积极塑造其AI未来的关键紧迫性。

英文摘要:

Africa's participation in modern AI development is constrained by severe infrastructural and policy gaps. Important barriers include limited access to high-performance computing (HPC), restricted cloud access due to payment system mismatches, volatile exchange rates, and strict data sovereignty laws that fragment regional collaboration between African Union (AU) member states. Although initiatives such as Cassava AI's network of AI factories to be deployed across the continent signal the growing interest in adopting AI in Africa, these projects remain very targeted, while continental adoption still requires better coordination between African stakeholders. Drawing on official declarations on AI adoption across the continent, this paper offers both qualitative and quantitative evidence that sustainable AI adoption requires robust digital foundations through balanced access to compute, data, and the energy that makes it possible. We refer to these foundations as the "right enablers", considering them as crucial components for success within the current context of the global AI race. We also introduce the \textit{Africa AI Compute Tracker (ACT)}, an interactive map to monitor the availability of AI-ready HPC systems throughout the continent. This tool represents the first open-source effort to consolidate data on Africa's evolving HPC landscape, and aims to encourage more transparency from local AI stakeholders while facilitating broader access for AI developers. The work presented in this paper underscores the urgency of tangible actions aimed at closing the AI divide and allowing Africa to actively shape its AI future.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑