The fourth paradigm unites Inspur business machine

  • Detail

On September 10, the fourth paradigm and Inspur jointly launched the AI software and hardware integrated machine product "prophet AIO" in Beijing. "AIO" means to create a full closed loop of "AI in one" Ai industrial application. The product design is based on the integration of the fourth paradigm prophet enterprise AI core system and Inspur openpower hardware architecture. At the same time, the two sides officially announced the establishment of the AI all-in-one machine joint laboratory to continuously empower the industry users with strong AI business support and development capabilities, and accelerate the implementation process of the AI industry

(fourth normal form × Inspur business machine strategy signing ceremony)

the fourth paradigm is an internationally leading AI technology enabler. The capabilities of its flagship product "the prophet of the fourth paradigm" cover the whole life cycle of AI application development, operation, management, etc; Inspur commercial machine provides openpower server based on open power technology. It is committed to creating a more complete and efficient overall solution for users, and provides professional and agile software and technical services with the characteristics of easy after-sales service. This joint venture, combined with the leading AI system of the fourth paradigm and the strong innovation ability of Inspur business machines in the field of computing architecture and hardware, will provide customers with more valuable intelligent computing products

out of the box integrated design makes accessing AI as simple as using it.

according to the Gartner enterprise AI case study report, more than 60% of China's large enterprises will develop their own AI solutions before 2022. In response to this, daiwenyuan, founder and CEO of the fourth paradigm, said that although every enterprise hopes to build its own core AI capabilities, the threshold is quite high, including server configuration, data processing, model parameter adjustment, model training, etc. He summarized the five necessary conditions "brain" in the technical specification JGJ 79 ⑵ 002 for ground treatment of AI buildings: big data, response, algorithm, computing resources and clear business needs 。 Cheng

prophet AIO is committed to solving the brain threshold problem, breaking the data threshold with the self-developed big data system, and establishing a standardized collection system for behavior data and feedback data; Automl (automatic machine learning) technology is used to break the algorithm threshold, so that business personnel who do not have AI foundation and computer skills can complete AI application through the prophet graphical interface; The hardware capability of Inspur commercial machines is used to supplement the last piece of puzzle in the computing resource layer, eliminating the configuration and debugging steps of the underlying AI. Daiwenyuan, the founder of the fourth paradigm, said at the scene: "prophet AIO has realized the complete coverage of the five elements of AI commercial landing, giving enterprises the ability to use AI out of the box. In the future, enterprises can access AI as easily as they can use it, just by plugging in, connecting to the network, starting up and so on."

leading integration of software and hardware architecture, achieving 10 times the performance improvement at the same cost

different from the exploration of AI software and hardware integration in vertical fields such as face recognition, prophet AIO is the first AI all-in-one product launched for large-scale data mining and machine learning computing, which can be used in risk anti fraud, competitive marketing, personalized recommendation, advertising computing, intelligent manufacturing It has been applied in multiple data mining and decision-making scenarios such as customer operation and product pricing, and has brought significant effect improvement - under the same cost, the overall performance of Prophet AIO is more than 10 times higher than that of ordinary servers

on the one hand, the performance improvement is due to the extreme performance of the prophet platform. As a high-dimensional machine learning platform, the prophet platform can support trillion dimensional feature processing at most, and a single node can support hundreds of millions of data storage and tens of thousands of real-time requests; On the other hand, it is due to the hardware infrastructure innovation, system optimization and parallel framework improvement of Inspur commercial machines. Vice president of Inspur Group Huleijun, general manager of Inspur Commercial Machinery Co., Ltd., said: "This cooperation is also a breakthrough attempt in the hardware field. For AI systems that can support trillions of high-dimensional data, such as the 'prophet platform', Power9 has a 7tb/s processor on-chip switching speed, 230 gb/s memory bandwidth, 25gb/s n its surface is generally not or rarely decorated with fabric or leather VLink and open capi interfaces, as well as the first batch of PCI-E 4.0 interfaces in the industry. The combination of these technologies enables it to accelerate data Transmission, which is 10 times faster than the x86 system based on PCI-E 3.0 in computing speed, can also significantly improve the performance of open-source AI frameworks such as chain, tensorflow and Caffe, and accelerate databases such as Kinetica, providing computing conditions for large-scale AI applications. In this way, data scientists can build applications including in-depth learning insight in scientific research, real-time fraud detection and credit risk analysis at a faster speed "

in the future, the competition of enterprises will be the competition of data scale and computing power. The AI software and hardware all-in-one product" prophet AIO ", with the technical advantages of AI application full process coverage and software and hardware integration design, It is expected to push AI technology into a dividend harvest period of large-scale implementation and great value release


Copyright © 2011 JIN SHI