Выпуск
The Hidden Physics That Could Make AI More Energy Efficient
The Quark Side - Quantum Physics Podcast · 31 октября 2026 г. · 19 мин
Researchers at the University of Innsbruck have uncovered a fundamental trade-off between energy efficiency and predictive accuracy in intelligent systems. Their findings suggest that AI agents may sometimes need to discard information about their past actions to maximize energy extraction. Governed by the laws of physics, this surprising limit could reshape how scientists design more energy-efficient and sustainable artificial intelligence. This episode includes AI-generated content.