As quantum computers continue to be researched for their potential to disrupt traditional hardware, there is also a growing interest in neuromorphic computers which offer a promising alternative to the binary system of 1s and 0s. ChatGPT 4, with over a trillion parameters, demonstrates the importance of scale in improving language models. However, the cost of running these models on conventional hardware is high and unsustainable, with millions of dollars required daily. The energy consumption of large language models like ChatGPT raises concerns, as they fall short of human intelligence efficiency.
Inspired by the human brain, neuromorphic computers use analog circuits to mimic brain functions and operate at low voltages, unlike traditional digital computers. Intel’s Hala Point, powered by Loihi 2 processors, is the world’s largest neuromorphic computer with over a billion artificial neurons and 128 billion synapses. It is significantly faster and more energy-efficient than conventional CPU/GPU configurations, performing 15 trillion 8-bit operations per second per watt.
While quantum computers are anticipated to bring about significant changes in hardware, the development of neuromorphic computers still requires further progress before becoming mainstream. Intel’s Hala Point, although currently a research prototype, demonstrates the potential for improving computational capabilities and sustainability in AI development. Access to Intel’s neuromorphic research communities allows for collaboration and advancement in this field, offering a glimpse into the future of hardware innovation.
Article Source
https://www.sify.com/technology/mainstream-ai-workloads-too-resource-hungry-try-hala-point-intels-largest-neuromorphic-computer/