The battle between the two major mobile phone AI chips: Qualcomm will increase investment and disdain Huawei's hardware to be launched first

Huawei's Mate 10, which is equipped with the industry's first artificial intelligence (AI) mobile chip, has recently made a lot of headlines, which has embarrassed Qualcomm, the leader in mobile chips. However, Qualcomm quickly responded, saying that it will increase investment in the AI ​​field in the next step.

At the Qualcomm 4G/5G Summit being held in Hong Kong, Keith Kressin, senior vice president of product management at Qualcomm, said in an interview with Tencent Technology, "Increasing investment in the field of AI will focus on improving the performance of CPU, GPU and DSP to better support AI."

He gave an example, "A few years ago, we added a Hexagon vector image to the DSP. It contains a huge 1024-bit library. Through this library, we can build some matrix expansions for AI and collect relevant data behind it. These data are important features of AI, so we have to increase investment in these three areas."

In order to speed up the execution of AI applications on smartphones, related companies are trying various possibilities. Obviously, unlike Huawei's hardware AI chip Kirin 970, Qualcomm focuses on AI's software capabilities.

In fact, not long ago, Qualcomm announced that the software development kit (SDK) of its Neural Processing Engine (NPE) has been launched to developers. The SDK was announced by Qualcomm last year and was only open to some partners at the time. Now all developers can use it.

Qualcomm said Facebook will be one of the first companies to integrate the SDK, and they are currently using it to speed up augmented reality filters in their mobile apps. By using the neural processing engine, Facebook's filters load five times faster.

This means that if you are developing an app that uses AI (such as image recognition), you can integrate Qualcomm's SDK to make it run faster on a compatible processor.

For Qualcomm, the next step of NPE is naturally hardware, that is, launching a dedicated AI mobile chip. As for when Qualcomm will release it, it is still unknown, but Qualcomm is not worried about Huawei's AI chip.

As the world's first AI mobile chip, Huawei's Kirin 970 has a built-in NPU (neural network processor) to enable chip-autonomous deep learning, which is different from Qualcomm's NPE.

The built-in NPU enables the Kirin 970 to process images 20 times faster than a single CPU (central processing unit). Huawei's internal test data shows that when used for image recognition, the Kirin 970 can process up to 2,005 images per minute, which is much faster than the 95 images/minute of the Samsung S8 CPU and the 487 images/minute of the iPhone 7 Plus.

But Qualcomm believes that AI is developing very fast and it is not the right time to talk about hardware at this stage. Keith Kressin said that it is not difficult to build AI unit modules into chips, and Qualcomm has done it before. The key lies in software development and optimization of programming modules.

"AI is developing and changing very quickly, so it requires hardware to be flexible and upgradeable rather than rigid. Therefore, we need to optimize and improve AI software to adapt to the ever-changing AI. Of course, both software and hardware will need to work together in the future, but Qualcomm is still focusing on the software level now," he said.

Regardless of the method, the user's understanding of AI is still based on the application experience, which is the key. From the perspective of competition, Huawei's lead in AI hardware has enabled AI to be implemented from software to hardware on the mobile phone side, which has had some impact on the status of Qualcomm chips. In the long run, it hopes to get rid of its dependence on Qualcomm chips as soon as possible.

At the same time, giants such as Samsung, Apple, Microsoft, and Google (Weibo) are also constantly strengthening their deployment of AI in their own ways. Obviously, some mobile reference standards are needed in the future, which is why each company is competing to provide the best solutions for AI deep learning.

Of course, there are huge business opportunities in the future. Regarding the industrial prospects of AI chips, some institutions predict that by 2025, the global AI chipset market will exceed 12.2 billion US dollars; and my country's "New Generation Artificial Intelligence Development Plan" predicts that by 2020, the domestic intelligent computing chip market will reach 10 billion yuan.

As a winner of Toutiao's Qingyun Plan and Baijiahao's Bai+ Plan, the 2019 Baidu Digital Author of the Year, the Baijiahao's Most Popular Author in the Technology Field, the 2019 Sogou Technology and Culture Author, and the 2021 Baijiahao Quarterly Influential Creator, he has won many awards, including the 2013 Sohu Best Industry Media Person, the 2015 China New Media Entrepreneurship Competition Beijing Third Place, the 2015 Guangmang Experience Award, the 2015 China New Media Entrepreneurship Competition Finals Third Place, and the 2018 Baidu Dynamic Annual Powerful Celebrity.