韩国研究人员开发双输出突触以实现高效AI

韩国大学的研究人员开发了一种双输出人工突触,以提升多任务AI系统的能量效率,该大学宣布。该设备同时发出电信号和光信号,实现并行处理。测试显示,计算速度提高高达47%,与传统GPU硬件相比,能耗降低最多32倍。

韩国大学KU-KIST融合科学技术的研究生院的研究人员,在王建旭和朴英兰教授的带领下,创建了一种受大脑启发的 artificial synapse。传统的AI芯片是为特定功能设计的,对于多个任务往往需要顺序处理,从而增加功耗。这种新设备通过同时发出电信号和光信号,在单个芯片上允许任务的并行处理。 团队报告在约1000个不同状态下实现了稳定的学习。在测试中,它实现了计算速度提高高达47%,与基于GPU的加速器相比,能耗降低高达32倍。 “这一成就通过同时利用电信号和光信号的人工突触,为多任务AI呈现了一种新的硬件架构,”王教授表示。“它可以进一步扩展到需要复杂决策的高速、低功耗AI系统,例如机器人、医疗保健应用和自动驾驶等领域。” 该研究于周五发表在《Science Advances》上,这是美国科学促进会(American Association for the Advancement of Science)的一本期刊。随着AI对计算能力和电力需求的增加,模仿大脑效率的努力正在获得更多关注。

相关文章

Illustration of Northwestern University's wireless micro-LED brain implant delivering light patterns to mouse neurons for sensory signaling.
AI 生成的图像

Northwestern team develops wireless implant that ‘speaks’ to the brain with light

由 AI 报道 AI 生成的图像 事实核查

Scientists at Northwestern University have created a soft, wireless brain implant that delivers patterned light directly to neurons, enabling mice to interpret these signals as meaningful cues without relying on sight, sound or touch. The fully implantable device uses an array of up to 64 micro-LEDs to generate complex activity patterns across the cortex, a development that could advance next-generation prosthetics and sensory therapies, according to Northwestern and Nature Neuroscience.

Researchers from Purdue University and the Georgia Institute of Technology have proposed a new computer architecture for AI models inspired by the human brain. This approach aims to address the energy-intensive 'memory wall' problem in current systems. The study, published in Frontiers in Science, highlights potential for more efficient AI in everyday devices.

由 AI 报道

Chinese researchers have introduced photonic AI chips that promise significant speed advantages in specific generative tasks. These chips use photons instead of electrons, enabling greater parallelism through optical interference. The development could mark a step forward in AI hardware, though claims are limited to narrowly defined applications.

Researchers have engineered a protein that detects subtle glutamate signals between neurons, unveiling a previously hidden aspect of brain communication. This tool allows real-time observation of how brain cells process incoming information, potentially advancing studies on learning, memory, and neurological disorders. The findings, published in Nature Methods, highlight a breakthrough in neuroscience.

由 AI 报道

A Los Angeles-based startup, Quilter, has developed a dual-PCB Linux computer using AI design, completing the project in one week with under 40 hours of human effort. The system, featuring 843 components, successfully booted on its initial try. This achievement highlights rapid AI-assisted hardware innovation.

Tesla Inc. is actively seeking AI chip designers in South Korea as part of its push to lead in AI chip production. The company announced the recruitment drive on Thursday, aiming to develop high-volume AI chip architecture. Applications are open for the AI Chip Design Engineer position.

由 AI 报道

Researchers have developed an ultrafast laser technique that fires light pulses in one billionth of a second, enabling the creation of structures 1,000 times stronger and 1,000 times faster. This novel method targets thermal conductivity in chips by controlling phonon scattering distances, offering applications in high-performance computing, quantum devices, and AI chip cooling. It changes how chips handle heat without relying on fans or liquid cooling.

 

 

 

此网站使用 cookie

我们使用 cookie 进行分析以改进我们的网站。阅读我们的 隐私政策 以获取更多信息。
拒绝