Un nuevo chip spintrónico promete mayor eficiencia en inteligencia artificial sin sacrificar precisión
The research presents a digital spintronic memory chip based on STT-MRAM that addresses the data movement bottleneck in artificial intelligence hardware. By integrating processing and storage within the same component, this architecture avoids energy-intensive transfers between the processing unit and memory. The digital approach ensures superior precision compared to traditional analog solutions, guaranteeing reliable results even when scaling to complex models, marking a significant advancement in hardware operational efficiency. This development has direct implications for the decentralization of artificial intelligence. By drastically reducing energy consumption and latency, the chip enables advanced AI tasks to be executed locally on portable devices, such as IoT sensors and mobile phones. This reduces dependence on data centers, improving user privacy and response speed. The ability to adapt precision according to need facilitates its implementation in demanding applications without sacrificing accuracy, paving the way for more sustainable and less polluting computing. The relevance to the field of open data lies in the fact that the democratization of AI hardware facilitates access to advanced processing tools in resource-constrained environments. When efficient, low-power computing becomes accessible, the barrier to entry is lowered for organizations and communities that need to analyze large volumes of information without relying on expensive cloud infrastructure. This fosters an ecosystem where the generation and analysis of open data can be carried out more agilely, privately, and ecologically, driving local innovation and data sovereignty.
Source: wwwhatsnew.comPublished on 2025-11-07