Turn off the Ad Banner  

To print: Select File and then Print from your browser's menu.

    -----------------------------------------------
This story was printed from CdrInfo.com,
located at http://www.cdrinfo.com.
-----------------------------------------------

Appeared on: Wednesday, July 4, 2018
Baidu Unveils High-Performance Kunlun AI Chip, AI Partnerships With Intel

Baidu Inc. today announced Kunlun, China's first cloud-to-edge AI chip, built to accommodate high performance requirements of a wide variety of AI scenarios. The company also announced partnerships with Intel.

The announcemenents were made at Baidu Create 2018 in Beijing, the second annual AI developer conference sponsored by China's Internet giant.

Kunlun

Kunlun is a high-performance and cost-effective solution for the high processing demands of AI. The chip can be applied to both cloud and edge scenarios, such as data centers, public clouds and autonomous vehicles. It leverages Baidu's AI ecosystem, which includes AI scenarios like search ranking and deep learning frameworks like PaddlePaddle.

In 2011, Baidu started developing an FPGA-based AI accelerator for deep learning and began using GPUs in datacenters. Kunlun, which is made up of thousands of small cores, has a computational capability of 260TOPS at 100W -- nearly 30 times faster than the original FPGA-based accelerator.

Other key specifications include: 14nm Samsung engineering, 512 GB/second memory bandwidth, as well as 260TOPS while consuming 100 Watts of power.

In addition to supporting the common open source deep learning algorithms, Kunlun chip can also support a wide variety of AI applications, including voice recognition, search ranking, natural language processing, autonomous driving and large-scale recommendations.

Mobileye & Baidu's Apollo

Intel/Mobileye announced that Mobileye's Responsibility Sensitive Safety (RSS) model will be designed into both Baidu's open-source Project Apollo and commercial Apollo Drive programs.

Baidu also announced plans to adopt Mobileye's Surround Computer Vision Kit as the visual perception solution; it will be integrated as part of Baidu's proposition to the Chinese OEM market.

Apollo is designed to be "an open, secure and reliable self-driving ecosystem" that can help members of the autonomous driving industry quickly build their own complete autonomous vehicle systems. Launched a year ago, it has enlisted 116 global partners, according to Baidu.

RSS is an open and transparent formal model that provides safety assurance of AV decision-making. RSS does this by formalizing common sense human-centered concepts of what it means to drive safely. Examples: always maintain a safe following distance and right-of-way is given but not taken.

The Mobileye Surround View Camera kit includes 12 cameras positioned around the vehicle plus Mobileye's computer vision (CV) hardware and software, which leverages the cameras combined view into a unified CV solution for autonomous cars.

The decision-making systems in autonomous vehicles today are based on artificial intelligence (AI) - meaning they operate probabilistically and can make mistakes. When it comes to safety and the lives of passengers, a probabilistic model is not enough. That's why Mobileye recommends adding a separate, deterministic layer on top of AI-based decision-making solutions in autonomous vehicles today.

Baidu and Mobileye will collaborate on the verification of RSS's formal model to the driving styles and road situations of the China market, and will jointly publish updates to the RSS model as this work results in new discoveries. "At Mobileye, safety assurance of automated vehicles is one of the most important issues facing the AV industry, and we are pleased Baidu has agreed to join us in this effort to deliver verifiable safety of AV decision-making into the China market," said Jack Weast, chief systems architect for Intel's Autonomous Driving Program.

RSS formalizes human notions of safe driving into a verifiable model with logically provable rules, defines appropriate responses, and ensures that only safe decisions are made by the automated vehicle and that the automated vehicle will do everything it can to avoid being involved in unsafe situations initiated by others.

Mobileye's RSS formulas will be integrated into the existing safety model (known as "DPS") of Baidu's Apollo Pilot. Apollo Pilot is Baidu's deployment version of Project Apollo and is being developed for multiple Chinese OEMs.

AI camera, FPGAs, PaddlePaddle

Separately, Intel used Baidu Create to showcase a host of its own AI-related technologies. These include Xeye, a new AI camera powered by Intel's Movidius vision processing unit, Baidu's plan to offer workload acceleration as a service which leverages Intel FPGAs, and Baidu's deep learning framework, called PaddlePaddle. Intel says Baidu's PaddlePaddle is now optimized for Intel Xeon scalable processors.

Baidu's Xeye camera uses Intel Movidius' Myriad 2 VPUs to deliver low-power, high-performance visual intelligence for retailers. Thanks to Intel's purpose-built VPU solutions coupled with Baidu's machine learning algorithms, the camera can analyze objects and gestures, while also detecting people to provide personalized shopping experiences in retail settings.

Baidu is developing a heterogeneous computing platform based on Intel's latest field programmable gate arrays (FPGA) technology. Intel FPGAs will accelerate performance and energy efficiency, add flexibility to data center workloads, and enable workload acceleration as a service on Baidu Cloud.

Baidu?s also accelerated production of its Apolong fully self-driving bus, developed in tandem with local manufacturer King Long and the first product to emerge from its open Apollo platform. It intends to ship the first 14-seater mini-buses to Japan by early 2019 in partnership with a SoftBank Group Corp. subsidiary. The Chinese company has cranked out 100 of the vehicles from its plant in southern China, it said in a statement.



Home | News | All News | Reviews | Articles | Guides | Download | Expert Area | Forum | Site Info
Site best viewed at 1024x768+ - CDRINFO.COM 1998-2024 - All rights reserved -
Privacy policy - Contact Us .