Release of AI-optimized Kunlun Chip a Game Changer for Baidu

R. Bhattacharyya

Summary Bullets:

  • During its AI developer conference, Baidu made several announcements that demonstrate how it is moving the Chinese AI market forward, but the release of its Kunlun chip stands out as a key move that repositions it in the not only the Chinese market, but also globally
  • With Kunlun, Baidu joins the ranks of a select few companies that not only offer an AI platform that helps enterprises deploy AI-infused solutions, but that have also developed their own hardware to maximize AI processing.

Baidu is hot on the heels of the likes of Microsoft and Google. Although already known as an ambitious player in the AI realm, primarily in China, the search engine provider hasn’t managed to establish itself as a major force in the space, until now. Earlier this month, Baidu announced that it is bringing to market an AI-optimized chip, called Kunlun.  With the move, Baidu joins the ranks of a select few companies that not only offer an AI platform that helps enterprises deploy AI-infused solutions, but have also developed their own hardware to maximize AI processing. 

Kunlun is a vast mountain range in China featured in Chinese mythology. Symbolizing China’s breadth and strength, the Kunlun chip lives up to its namesake. It is almost 30 times faster than Baidu’s FPGA-based AI accelerator, offering 260 tera-operations per second and 512 GB/second memory bandwidth. The chip can be used to provide AI capabilities such as speech and text analytics, natural language processing, and visual recognition, as well as to support deep learning via Baidu’s PaddlePaddle platform. Available in two models, the Kunlun 818-300 is optimized for training complex machine learning and deep learning algorithms, whereas the Kunlun 818-100 is designed for inference.

As with competing hardware, the chips can be deployed in the cloud or within a data center for use in AI-related processing. The Kunlun chip can also be deployed at the edge, such as in autonomous vehicles, an area in which Chinese companies are allocating sizeable research and development funds. In fact, at its developer conference, Baidu showcased the Apolong, a mass-produced self-driving bus that uses its Apollo autonomous driving system. But edge deployments of AI don’t stop with autonomous vehicles, and the trend is gaining traction. On-device AI is used in mobile phone cameras to improve picture quality via visual recognition, it can provide speech and voice recognition, and it may be used in security systems, drones, or robots.  AI at the edge can increase efficiency since at least a portion of the analysis, if not all, can be performed without the need to transport data to and from the cloud, reducing latency.  It also offers greater flexibility, since the device can utilize AI even when it is offline, and it can improve the user experience since the device can learn behavior patterns.  Some users may prefer it since data stays on the device instead of being transmitted over a network to the cloud, and see it as a privacy-related benefit.

The release of the new chipset underscores the overall momentum behind AI in China, as well as the determination of Chinese players to establish themselves as global leaders in this emerging area. However, can a Chinese-focused service provider become a globally recognized provider of AI solutions and hardware? Baidu doesn’t market heavily to other regions, and will have a tough time competing with the likes of Amazon, Google, and Microsoft, who are well-entrenched among enterprises operating throughout the globe. Nonetheless, the company’s recent efforts demonstrate its aggressive approach, and underscore what we already know: that Chinese players have bet the farm on AI, and that we should expect to see more from them in the near future – much, much more.

 

What do you think?

Please log in using one of these methods to post your comment:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.