【英特尔发布全球最大神经拟态系统，性能超人脑200倍】
美国芯片巨头英特尔近日宣布了一项重大突破，成功构建了全球最大的神经拟态计算系统——Hala Point。这一系统旨在推进未来类脑人工智能的研究，以应对AI在效率和可持续性上的挑战。据钛媒体报道，Hala Point内置了1152个基于Intel 4（7nm）制程的Loihi 2处理器，其强大的计算能力可模拟11.5亿个神经元和1280亿个突触，每秒能处理超过380万亿个8位突触操作和240万亿个神经元操作。
Hala Point的独特之处在于，当应用于仿生尖峰神经网络模型时，即使在神经元数量相对较低的情况下，其运算速度也能达到人脑的200倍，即20000%的速度提升。这一成就为人工智能的发展开辟了新的可能，尤其是在需要高效、低能耗解决方案的领域。
该系统已初步部署在美国桑迪亚国家实验室，预计将进一步推动神经拟态计算在能源效率、机器学习和复杂决策系统等领域的应用。英特尔的这一创新标志着人工智能硬件的进步，为未来智能系统的性能提升和能源效率优化奠定了坚实基础。
英语如下：
**News Title:** “Intel Unveils Hala Point: World’s Largest Neuromorphic System, Outperforms Human Brain by 200 Times!”
**Keywords:** Intel, Hala Point, Neuromorphic System
**News Content:**
Intel, the American chip giant, recently announced a major breakthrough with the successful construction of the world’s largest neuromorphic computing system, known as Hala Point. Designed to advance brain-inspired AI research and tackle efficiency and sustainability challenges in AI, this system has been reported by Titanium Media.
Equipped with 1152 Intel 4 (7nm) based Loihi 2 processors, Hala Point boasts the capability to emulate 11.5 billion neurons and 1280 billion synapses, processing over 380 trillion 8-bit synaptic operations and 240 trillion neuron operations per second.
The uniqueness of Hala Point lies in its performance when applied to bio-inspired spiking neural network models. Despite having a relatively lower number of neurons, it operates at 200 times the speed of the human brain, representing a 20,000% increase. This milestone opens up new possibilities for AI development, particularly in areas demanding efficient and low-power solutions.
The system is initially deployed at the Sandia National Laboratories in the United States, with expectations that it will propel the application of neuromorphic computing in energy efficiency, machine learning, and complex decision-making systems. Intel’s innovation signals a step forward in AI hardware, laying a robust foundation for future improvements in system performance and energy efficiency.
【来源】https://www.tmtpost.com/7044204.html
Views: 1