编辑 | 学术君
▲天机芯片单片(左)和5×5阵列扩展板(右)
▲自动行驶自行车演示平台
Nature封面文章 | 中国团队的“天机芯”架起了通用人工智能领域机器学习和神经科学的桥梁
There are two main approaches to developing artificial general intelligence. One is rooted in neuroscience, and attempts to construct circuits that closely mimic the brain. The other is grounded in computer science, and uses computers to execute machine-learning algorithms.
然而,由于算法和模型的差别,当前人工智能芯片均只支持人工神经网络或脉冲神经网络,难以发挥计算机和神经科学两个领域的交叉优势。
In this week’s issue, Luping Shi and his colleagues reveal the Tianjic chip — an electronic chip that integrates the two approaches into one hybrid platform. The Tianjic chip has multiple functional cores that are readily reconfigurable, enabling it to accommodate both machine-learning algorithms and brain-inspired circuits.
A key innovation from the research team is Tianjic’s unified function core (FCore) which combines essential building blocks for both artificial neural networks and biologically networks — axon, synapse, dendrite and soma blocks.
The 28-nm chip consists of 156 FCores, containing approximately 40,000 neurons and 10 million synapses in an area of 3.8×3.8 mm².
The research team also showcased the superior performance of Tianjic compared to GPU, where the new chip achieves 1.6–100 times better throughput and 12–10000 times better power efficiency.
研究人员将芯片安装到无人自行车,来证明该芯片的强大功能和潜力。这款无人自行车能自我平衡,具有声控功能,可探测并避开障碍物。这都是“天机”芯片能同时处理多种算法和模型的强大功能。
The researchers demonstrate the potential of this approach by incorporating one of their chips into a riderless autonomous bicycle, which can self-balance, is voice controllable and can detect and avoid obstacles, all as a result of the Tianjic chip’s simultaneous processing of versatile algorithms and models.
Equipped with the Tianjic chip and IMU sensor, a camera, steering motor, driving motor, speed motor and battery, the bicycle was tasked with performing functions such as real-time object detection, tracking, voice-command recognition, riding over a speed bump, obstacle avoidance, balance control and decision making.
The research team developed a variety of neural networks (CNN, CANN, SNN and MLP networks) to enable each task. The models were pretrained and programmed onto the Tianjic chip, which can process the models in parallel and enable seamless on-chip communication across different models.
The research team told Chinese media they expect the Tianjic chip to be deployed in autonomous vehicles and smart robots. They have already started research on the next-generation chips and expect to close the R&D stage early next year.
论文链接:
https://nature.com/articles/s41586-019-1424-8
本文来源:中国日报、清华大学官微、募格学术
转载本文请联系原作者获取授权,同时请注明本文来源。
如果不想错过每日精彩
一定要星标置顶哟~▼
▼ 点击「阅读原文」,获取更多免费资源
本篇文章来源于微信公众号: 学术资源大全
上一篇: 今日最佳:某高校的硬核刷脸机
下一篇: 今日最佳:新生入学秀出的一波硬核操作
相关文章
版权声明:
本网站(网站地址)刊载的所有内容,包括文字、图片、音频、视频、软件、程序、以及网页版式设计等均在网上搜集。
访问者可将本网站提供的内容或服务用于个人学习、研究或欣赏,以及其他非商业性或非盈利性用途,但同时应遵守著作权法及其他相关法律的规定,不得侵犯本网站及相关权利人的合法权利。除此以外,将本网站任何内容或服务用于其他用途时,须征得本网站及相关权利人的书面许可,并支付报酬。
本网站内容原作者如不愿意在本网站刊登内容,请及时通知本站,予以删除。