Heat-induced electrical fluctuations treated as probabilistic resource rather than as error, to power and improve AI

https://techxplore.com/news/2026-03-thermal-noise-framework-power-ai.html

"AI/ machine learning inherently probabilistic... uses thermal noise's natural randomness to drive probabilistic bits enabling much more efficient AI sampling/ optimization... using existing thermal noise to perform calculations drastically reduces power required... harnesses ambient heat to run AI algorithms... faster generative AI/ Boltzmann machines... nanomagnetic/ spintronic device compatible... AI-specific hardware orders magnitude smaller/ more energy-efficient than current GPUs... edge devices, wearables"

Related:

Brain-inspired device could lead to faster, more energy-efficient AI hardware
https://www.eurekalert.org/news-releases/1119548

Introducing LightGen, a chip for ultra-fast, ultra-efficient generative AI
https://www.eurekalert.org/news-releases/1109809

KAIST Researchers Develop Brain-Inspired AI Model That Improves Learning Stability
https://www.kaist.ac.kr/newsen/html/news/?mode=V&mng_no=58990

New photonic chip runs AI in trillionths using light, cuts heat and energy use
https://interestingengineering.com/energy/photonic-ai-chip-university-sydney