The current era of artificial intelligence (AI) is characterized by the scaling of data and computational resources as the primary driver of emergent capabilities in AI models. However, this trend faces fundamental constraints on data availability and computing power. Recent breakthroughs suggest an alternative path forward—leveraging novel statistical and information-theoretic tools to enhance reliability and data efficiency, whileÌý enhancing intelligence per joule and per data point/token via hardware-software co-design principles. Understanding the fundamental limits of AI through the lens of information and physical principles, such as Landauer's principle, is crucial for developing sustainable and efficient learning systems. This special issue aims to advance theoretical and algorithmically motivated approaches to optimizing AI performance while reducing reliance on extensive data and energy resources.
Guest Editors
Bipin Rajendran (King’s College London) – Lead guest editor
Osvaldo Simeone (King’s College London) – Lead guest editor ÌýÌý
Irem Boybat (IBM) ÌýÌý
Tianyi Chen (Rensselaer Polytechnic Institute) ÌýÌý
Siddarth Garg (NYU) ÌýÌý
Yaniv Romano (Technion)
Manuscript Submission Deadline: Continuous submission untilÌýNovember 15, 2025
Publication Date: Each accepted manuscript will be published on ½ÄÏÌåÓý Xplore after finishing its peer-review with a final deadline for publishing the complete special issue by June 1, 2026
More information:Ìý/jsait/jsait-call-papers/special-issue-energy-and-data-efficiency-artificial-intelligence