The global smart phone market as a whole suffered winter, and the Chinese market was the hardest hit. However, almost all headphone companies have achieved significant growth. In the fourth quarter of 2017, the market share of the top five brands in the Chinese mobile phone market has exceeded 80%. It is expected that in 2018 this figure will be close to 90%.
The biggest surprise is that there are still variables in the giant market. For example, Huawei’s handset shipments have exceeded Apple’s for two consecutive months in mid-2017, and the overall market share gap between the two is about 4 percentage points in 2017, but Huawei’s shipments are 153 million units per year. It still maintains a rapid growth of nearly 10%. It is expected that Huawei's mobile phone shipments in 2018 will surpass Apple's number one smartphone company in the world.
However, in the face of major technological changes, compared with the shipment figures, the competition between the two giants' mobile phone AI chips and the ecology has attracted more attention in the industry.
What is the difference between AI mobile phone giants?
Smartphones have 'flattened' the entire technology industry. They have also allowed Apple and Huawei, two almost identical companies, to start a positive competition.
In 2017, Huawei announced the world's first artificial smartphone processor, the Unicorn 970, and subsequently introduced the chip-grade artificial smartphone Huawei Mate 10. Apple also released the iPhone X with the A11 Bionic processor after Huawei.
Different from Qualcomm's traditional chip enterprise adopting a distributed architecture artificial intelligence processor solution, Huawei and Apple artificial intelligence processors have adopted a more aggressive deep customization (ASIC) scheme, although the latter has a longer design time and a higher threshold. , But the advantage is that through a dedicated processing unit, the efficiency and speed of invoking a distributed architecture scheme compared to software distribution will increase by 5 to 10 times, and the energy consumption control will have an advantage, which is why Huawei and Apple coincide at this point. s reason.
However, Apple and Huawei's first artificial intelligence processor has a clear difference in application focus: The Apple A11 Bionic is mainly used to enhance iPhone X's own portrait recognition, such as Face ID, photo scene recognition, and Animoji animated expressions. Xuan Ji' composition, the current application of a narrow range. Huawei is more pragmatic, Kirin 970's artificial intelligence is mainly used in mobile phone photography AI eye eye-witness, intelligent voice and real-time translation, information services such as direct access to more scenes and practical wisdom services.
The different characteristics of Apple's A11 Bionic and Qilin 970 show that from the three-year or even earlier definition stage of the processor product, the two giants of Apple and Huawei have obvious differences in the positioning and ecological layout strategies of artificial intelligence terminals.
Ecological battle of artificial smart phone chips
From the PC Internet to the mobile Internet, the giants have won over platforms and ecological disputes, such as Microsoft PCs vs. Apple Macintosh, Google and Apple vs. Nokia in the feature-plane era, as well as the era of artificial intelligence.
Apple introduced the machine learning framework Core ML to developers for artificial intelligence applications, which reduced the difficulty for developers to develop AI applications. The problem is that Core ML is not open source, and must be based on Apple's own equipment and ecological development. The strategy is relatively closed. In the Apple A11 Bionic open AI capacity is limited, iPhone X sales are limited, artificial intelligence experience is still weak, no doubt increased the developer's input costs and risks.
In contrast, Huawei has learned the experience and lessons of the ecological competition between Apple and Google in the era of mobile Internet: Launched Unicorn 970 with a separate artificial intelligence chip. At the same time, it also launched a comprehensive AI open platform based on HiAI architecture to support Tensorflow. Major global neural network programming frameworks such as /Tensorflow Lite and Caffe/Caffe 2. Compared to Apple Core ML, the HiAI architecture can not only leverage the integrated experience of Huawei's artificial intelligence chip-operating system-mobile phone hardware-terminal cloud service, but also rely on it. Open strategy to minimize development thresholds and costs for developers in AI applications.
Internet giants also compete to launch their own AI development kits, but for ordinary developers, they must purchase their cloud services. AI capabilities are based on the combination of cloud computing and big data, or according to the flow of fees, test deployment will Generate costs. Collecting users' use of big data will incur costs. The growth of bandwidth and traffic caused by user growth will also incur costs. Obviously, small development teams and individual developers cannot afford it. Large and medium-sized development companies cannot afford strategic binding and future. Uncertainty, while still having to face different end-side adaptations.
From this point of view, the HiAI architecture is currently the only open platform that provides cloud-side computing + end-to-end AI capabilities. Based on the special AI processing capabilities of the NPU in the Unicorn 970 chip, it can provide developers with more than 25 times the CPU and 10 times the GPU. , With AI calculations with 50x energy efficiency, you can avoid high cost pressures, save a lot of development costs, realize short-term development and gain, and continuously optimize and optimize your AI application experience. For example: In 2012, Wu Enda had 1 billion nodes in the neural network established by the leading Google Brain Cat Face Recognition Project. It used 16,000 CPUs in total and took 7 days. Now Huawei Hass 970 processor with NPU is ready for use. Real-time recognition of 13 objects including cat faces and photo scenes.
At this stage, whether the Internet companies of Apple iPhone X, Baidu, Ali, or other Internet companies have achieved their experience using separate AI capabilities, such as A11 Bionic's neural network processing engine can recognize and deeply learn the facial features of people, as Face ID, Portrait Lighting provides better performance support, Alibaba's Shopping for Ideas, Baidu DuerOS intelligent speech interaction, and semantic deep learning programs. As a result, the developer's breakthrough in this direction may not be large. The platform subdivides AI technology to improve existing mobile applications.
In response to this industry dilemma, HiAI architecture provides video (face recognition, gesture recognition, portrait segmentation, human pose recognition, etc.), photographing (beauty, image enhancement, AI recognition), AR (environmental understanding, SLAM), shopping (Graphic shopping), social networking (photo classification, image recognition), translation (photographing translation, speech recognition and translation) and other six AI capabilities and solutions, covering almost all major application areas of the current mobile Internet. Developers can pass The HiAI architecture invokes the NPU to improve existing mobile applications, and can also create new AI innovation applications by associating several AI capabilities.
More importantly, the HiAI architecture uses AI as a universal basic capability. By supporting the global mainstream neural network programming framework and open application layer APIs, developers can focus more on the product innovation of AI applications, even if they do not understand AI algorithms. Developers can also complete high-quality application development and experience optimization in a short period of time.
From this point of view, at this stage, cloud-testing AI is mainly applied to the validation of AI capabilities between giants and the optimization of their own existing services. End-point AI is more biased toward end-user experience and developer innovation. Apple also provides an open-ended AI platform. Huawei, the former is still a closed ecological road in the era of mobile Internet. Huawei chose a comprehensive open strategy. The battle between the 'empire' and 'republic' ecology of the artificial smart phone chips will also affect the AI wisdom. Cellular future competition and industry trends.
The mobile phone industry goes to Huawei's best technical dispute
Huawei is a typical technology-driven enterprise. In the past 10 years, Huawei's research and development investment has reached 310 billion yuan. In recent years, Huawei's investment in R&D expenses has risen to about 15% of its revenue. In 2016 alone, it invested 76.4 billion yuan. In 2017, R&D investment was as high as 10.7 billion euros (approximately 8.11 billion yuan), surpassing Apple (9.5 billion euros), ranking sixth in the world, and the only Chinese company in the world's top 50 in R&D spending. Hassell Kirin Chips It is Huawei's insistence on a huge amount of R&D and staffing core products for nearly 14 years.
In the past 30 years, Huawei has defeated Nokia, Siemens, Alcatel, Lucent, Ericsson and other global telecommunications giants in succession, the employees’ will to fight, the heavy technology investment mode to achieve breakthroughs with points, and the close combat in the global campaign. , The international giants are often impressed.
This is also true in the mobile phone industry. Huawei is the world’s third-largest smart phone company. From the functional machine to the smart phone, the PC Internet has transformed to the mobile Internet. Huawei has also realized the importance of the platform and the ecological dispute. It faces the new 5G. , Artificial intelligence, the key technological change of the smart Internet, actually returned to Huawei's most familiar technical route, the world's first artificial intelligence processor overseas Kirin 970 release allows Huawei to open up with Apple's positive 'hard' ', This is also a game that Huawei is good at.
According to the laws of the chip industry and Huawei, Apple’s product technology line, the two mobile phone giants will release the second-generation mobile phone AI processors this year. Huawei's advantage lies in the annual Mate and P series, and the glory digital series and V series. More product iterations and innovations, as well as the HiAI architecture's comprehensive AI open strategy appeal to a large global developer community. This will be a race against time, production volume competition, eco-strategy winning all-round contest, and also the 2018 mobile phone. One of the industry's biggest insights and variables.