Alibaba is making its own neural network chip
Apr 20, 2018Masha Borak
AlibabaArtificial IntelligenceNews
Alibaba is developing its own neural network chip, the Ali-NPU, which will be used in AI applications, such as image video analysis, machine learning, and other scenarios, Yicai is reporting. After the chip is mature, it will provide services for businesses through Ali Cloud.
But this isn’t the only news—Alibaba claims that the chip’s performance to price ratio will be 40 times that of current products on the market. According to the company, the Ali-NPU chip will have a special architecture with 10 times better performance than mainstream AI chips based on CPU or GPU architecture, while the manufacturing cost and power consumption are only half of that. Some commentators have questioned if this is a valid comparison since CPU and GPU chips have different architectures than NPUs.
The chip is being developed by Alibaba’s R&D institute, the DAMO Academy which stands for “Discovery, Adventure, Momentum, and Outlook.” The Academy was founded in October 2017 as part of Alibaba’s wider push into cutting-edge tech and it plans to attract talent by opening centers in seven cities around the world.
Last month during Alibaba’s inaugural tech summit called “New Technology, New Future”, the e-commerce company announced that it would be investing in technological infrastructure including machine learning, chips, the Internet of Things, operating systems, and biometric identification. The event was held at Hangzhou headquarters on March 9 and was attended by around 5,000 engineers.
“An economy that serves two billion people must be backed by solid technological capacity. To shoulder the future responsibility, we will build Alibaba’s own ‘NASA,’” Alibaba’s executive chairman Jack Ma said during the event.
AI chips are one of the main targets of China’s push for AI development. Currently, there are dozens of people working on chip development at the DAMO academy in Shanghai and the US while that number is set to reach 100, Yicai reports. Alibaba has invested in several chip companies around the world, including China-based Cambricon (寒武纪), Kneron (耐能), ASR (翱捷科技), C-Sky (中天微), and DeePhi (深鉴, as well as California-based Barefoot Networks.
Among Alibaba’s new technology focuses are also self-driving vehicles and quantum computing where it aims to compete with local rival Baidu.
Apr 20, 2018Masha Borak
AlibabaArtificial IntelligenceNews
Alibaba is developing its own neural network chip, the Ali-NPU, which will be used in AI applications, such as image video analysis, machine learning, and other scenarios, Yicai is reporting. After the chip is mature, it will provide services for businesses through Ali Cloud.
But this isn’t the only news—Alibaba claims that the chip’s performance to price ratio will be 40 times that of current products on the market. According to the company, the Ali-NPU chip will have a special architecture with 10 times better performance than mainstream AI chips based on CPU or GPU architecture, while the manufacturing cost and power consumption are only half of that. Some commentators have questioned if this is a valid comparison since CPU and GPU chips have different architectures than NPUs.
The chip is being developed by Alibaba’s R&D institute, the DAMO Academy which stands for “Discovery, Adventure, Momentum, and Outlook.” The Academy was founded in October 2017 as part of Alibaba’s wider push into cutting-edge tech and it plans to attract talent by opening centers in seven cities around the world.
Last month during Alibaba’s inaugural tech summit called “New Technology, New Future”, the e-commerce company announced that it would be investing in technological infrastructure including machine learning, chips, the Internet of Things, operating systems, and biometric identification. The event was held at Hangzhou headquarters on March 9 and was attended by around 5,000 engineers.
“An economy that serves two billion people must be backed by solid technological capacity. To shoulder the future responsibility, we will build Alibaba’s own ‘NASA,’” Alibaba’s executive chairman Jack Ma said during the event.
AI chips are one of the main targets of China’s push for AI development. Currently, there are dozens of people working on chip development at the DAMO academy in Shanghai and the US while that number is set to reach 100, Yicai reports. Alibaba has invested in several chip companies around the world, including China-based Cambricon (寒武纪), Kneron (耐能), ASR (翱捷科技), C-Sky (中天微), and DeePhi (深鉴, as well as California-based Barefoot Networks.
Among Alibaba’s new technology focuses are also self-driving vehicles and quantum computing where it aims to compete with local rival Baidu.
Code:
https://technode.com/2018/04/20/alibaba-npu/