Baidu’s Big AI Rollout Includes a Partnership With Intel
BEIJING, CHINA – AI is well known as a hot area of innovation from the likes of Google, IBM, Microsoft and – Baidu? You may not have heard of the Chinese tech giant but it’s starting to make waves across a range of product areas including advanced chips, robotics, autonomous vehicles and artificial intelligence.
At its annual Baidu Create conference here, Baidu announced a partnership with Intel, an advanced chip architecture of its own design and a new version 3.0 of its Baidu Brain AI software.
As part of the Intel partnership, Baidu announced Xeye, a new camera designed to help retailers offer a more personalized shopping experience. The camera uses Baidu’s advanced machine learning algorithms to analyze objects and gestures as well as detect people in the store. It also leverages Intel’s Movidius vision processing units (VPUs) to give retailers low-power, high performance “visual intelligence” as to products and activity in the store.
Separately, Baidu is improving machine vision performance via its EasyDL, an easy-to-use “code free” platform designed to let users build custom computer vision models with a simple drag-and-drop interface. Released in November as part of Baidu Brain 2.0, EasyDL applications are being used by 160 grocery stores in the U.S. including Price Chopper. The computer vision application recognizes items left in a customer’s shopping cart by mistake to help ensure that they’re purchased.
The newer Baidu Brain 3.0 makes it easier and quicker to train a computer vision model using EasyDL so, for example, the application designed for the grocery cart can now be developed in as little as 15 minutes.
In addition to Xeye, Baidu also announced it will use Intel’s FPGAs (Field Programmable Gate Arrays) to enable workload acceleration as a service on the Baidu Cloud. “The best is yet to come. We are excited to see the innovative Baidu Brain running on Intel Xeon processors,” said Gadi Singer, general manager of Intel’s AI Products Group who joined Baidu CEO Robin Li on stage.
But Baidu has big chip plans in its own right. During his keynote, Li announced Kunlun, China’s first cloud-to-edge AI chip, designed for high performance AI scenarios. Li said Kunlun will be marketed for use in data centers, public clouds and autonomous vehicles.
Baidu started developing an FPGA-based AI accelerator for deep learning in 2011 and began using GPUs in datacenters. Kunlun, which is made up of thousands of small cores, has a computational capability which is nearly 30 times faster than the original FPGA-based accelerator.
And while the initial market for Kunlun will be China, Technalysis Research analyst Bob O’Donnell said enterprises across the globe would be wise to be aware of Baidu’s growing product portfolio.
“Baidu is a key player for multinational corporations with a presence in China because they’re driving innovation in the same way that Amazon or Google is in the U.S.,” said O’Donnell. “They have an incredibly strong focus on AI across a lot of different industries that’s as broad as any other company I know of. Right now they’re very China-focused, but I expect that to expand over time.”
Chip rivals like Nvidia have made huge strides in support of autonomous vehicles with both hardware and software frameworks and simulation software for testing designed to help car makers get vehicles to market.
Similarly, Baidu has made a big investment in its Apollo software for autonomous vehicles of all sizes, from automated wheelchairs to cars, buses, trucks and other transport vehicles. At Create it showed off the new Apollo 3.0 software that is just starting to be used in autonomous vehicles in campuses and other closed environments in China such as senior living communities.
“We are really excited, this will surely change everyone’s lives,” said Li, who announced the 100th autonomous buses had recently come off the assembly line.
“You can see this is a real automatic driving solution, there’s no steering wheel, brake pedal or throttle, but it’s also very stylish inside,” said Li.
The vehicles are planned for commercial use in both China and Japan.
Analyst O’Donnell said it looks like Baidu’s autonomous vehicle effort is focused on the Asian market for at least the near term. “But they’re really establishing some important benchmarks here with the breadth of what they’re doing that competitors are sure to take note of.”