The advancement of technology and artificial intelligence (AI) are proceeding at an ever-accelerating pace, raising important questions regarding the sustainability of this development.
One of the most innovative responses to this challenge comes from Intel, a giant in the field of chip manufacturing, which recently announced the creation of the largest neuromorphic system in the world: HalaPoint.
Intel has taken a significant step towards a sustainable future for artificial intelligence with the construction of HalaPoint. This system, initially deployed at Sandia National Laboratories, represents a milestone in the field of AI thanks to the use of Intel’s LoiHe2 processor. The main focus of this initiative is to support research for artificial intelligence inspired by the functioning of the human brain and address the challenges related to the efficiency and sustainability of current AI.
Mike Davies, director of the Neuromorphic Computing Lab at Intel Labs, highlighted how the industry needs fundamentally new approaches that can scale effectively. From this need, HalaPoint was born, a system that combines the efficiency of deep learning with new learning skills and brain-inspired optimization.
Unique Capabilities for the Future of AI
HalaPoint doesn’t just deliver cutting-edge computational efficiencies on mainstream AI workloads. The system also boasts unique capabilities that could pave the way for future AI-capable applications continuously learn in real time. This feature has the potential to revolutionize various industries, including scientific and engineering problem solving, logistics, smart city infrastructure management, large language models, and AI agents.
Craig Vinyard, team lead for HalaPoint at Sandia National Laboratories, shared how research with a system of this scale will keep pace with the evolution of AI in myriad fields from commercial to defense to basic science. The intent is to use HalaPoint for advanced research in brain-scale computing focusing on solving scientific computational problems in device physics, computer architecture and computer science.
Currently a prototype for advanced research in large-scale neuromorphic computing in the brain, the experiences gained with HalaPoint are expected to lead to viable advancements in future commercial systems. In particular, we hope to enable the great linguistic models to continuously learn from new data without interruption.
With up to 1.15 billion artificial neurons and 128 billion synapses distributed across 140544 neuromorphic processing cores, consuming a maximum of 2600 watts, this system demonstrates how it is possible to push the current boundaries of AI technology beyond while keeping a careful eye on energy sustainability.
Intel through the development of its innovative neuromorphic system; it demonstrates not only a solid response to the challenges presented by today’s artificial intelligence but also a promising vision towards revolutionary future applications spanning numerous fields of application.
