Home > News content

China AI Mobile Industry Report 2018: AI Chips and AI Algorithms have become the key

via:CnBeta     time:2018/12/19 18:10:09     readed:488

The report points out that AI mobile phones have entered an early stage of development, and some functions have played a decisive role. According to Erie consultation analysts, AI has become a new driving force for the growth of mobile phone market. In the future, upstream and downstream manufacturers will further improve the industrial chain layout through in-depth cooperation. Among them, AI head enterprises will play a more important role in the industrial structure and business model of AI mobile phones, and provide a sustained impetus for the development of the entire industry.

AIUpstream Mobile Industry: AI Chip and AI Algorithms are the Key

From the perspective of AI mobile phone industry structure, the upstream of AI mobile phone industry chain consists of "hard base" suppliers (mainly referring to AI chip suppliers) and "soft base" suppliers (mainly referring to AI algorithm suppliers). The difference with the traditional mobile phone industry chain is that the new AI architecture designer and the traditional chip manufacturer cooperate to provide AI chips, and the integration link introduces the AI algorithm provider.

AI chip has become an important criterion for judging AI mobile phones. At present, mobile chip giant manufacturers have brought different AI Implementation Paths to the industry. Qualcomm uses NPE to schedule CPU, GPU and DSP in its SoC, aiming to create a balanced and widely adapted AI solution. And Apple, Lianfeng, Samsung and HefeiHUAWEIHess has developed independent AI processor modules to accelerate AI applications. According to the report, there is no significant difference in the effectiveness of these two approaches.

AI algorithm in mobile phones can be run in two ways: terminal and cloud, among which terminal operation is the mainstream. There are two schemes for terminal AI, one is to run the algorithm on SoC and the other is to run the coprocessor on local components. Algorithms in SoC can be implemented by calling general hardware units to implement a variety of AI applications, or by adding independent processor units, which is the most mature and most widely used solution at present; AI processing has been implemented in coprocessors on local components and has been commercialized. At present, there are only cases of intelligent image processing. Algorithms mainly use the processing unit matrix to implement AI processing in SOC. The camera intelligently accelerates the photo processing and realizes the effect of single photo visualization. At present, Google explores and uses this scheme on its own mobile phone products.

In the main aspect of technology development, computer vision technology has become mature, reaching the key requirements of real-time and reliable landing on mobile phones; however, intelligent semantic interaction technology can only basically meet the needs of daily applications, and there is still much room for development.

AIMobile phone industry terminal and application: focus on ApplicationDevelopmentPerson and platform

From the terminal point of view, the cooperation between technology providers and mobile phone manufacturers has become increasingly rich and in-depth, gradually moving towards high-frequency scenarios, in addition, the cooperation mode has begun to mature.

In the aspect of mobile application of AI mobile phone, all kinds of players focus on the upstream Application Developer Platform in order to gain the ecological dominance of AI mobile phone and even IoT industry, such as Huawei HiAI, AR Engine, Huawei and Baidu Paddle Paddle Paddle to build cooperative ecosystem, Shangtang Technology's SENSEAR platform, Millet Mobile AI Compute Engine, etc. in the long run, they all strive for stronger voice in the industry standards. Meaning.

In landing applications, it mainly includes face unlocking, image processing and photo optimization, intelligent photo album, special effects of mutual entertainment, 3D display, intelligent voice assistant, casual translation, intelligent screen, system optimization and so on. Among them, from the user's point of view, the most valuable AI function is face unlocking, and the most commonly used function is voice assistant.

AIMobile phones will become the mainstream, consumers have greater expectations

AI began to promote the iteration of mobile phone products, and the market entered the stage of intelligent driving. The report predicts that the proportion of AI mobile phone shipments will increase from less than 10% in 2017 to 80% in 2022, and annual sales will exceed 1.3 billion units. AI mobile phones will be the direction of the next round of competition.

At the same time, the report argues that the AI mobile phone's title is to maximize the effectiveness of mobile phone software and hardware upgrades through AI technology, focusing on "upgrades" rather than "subversions", and focusing on such important application scenarios as vision and voice.

According to the report, 95.9% of Chinese mobile phone users will choose to buy AI mobile phones the next time they change their mobile phones. Consumers are enthusiastic about AI mobile phones and hope to make life more convenient through AI. In the survey of "Chinese mobile phone users'expectations for the future of AI mobile phone functions in 2018", Chinese mobile phone users put forward more demands for many AI functions, such as intelligent voice assistant, intelligent optimization, AI photography, intelligent recognition, unlocking mode, etc.

The continuous upgrading of user needs requires not only the technology provider to maintain the continuous iteration of the algorithm, but also its rapid innovation for the landing of scenario applications. Wang Xiaogang, co-founder of Shangtang Science and Technology and Dean of the research institute, pointed out that "the core value of AI mobile phones lies in self-learning, self-service, active understanding of user needs, and even becoming the"brain"to help users make decisions.

Download link for complete report: http://report.iresearch.cn/report_pdf.aspx?Id=3312

China IT News APP

Download China IT News APP

Please rate this news

The average score will be displayed after you score.

Post comment

Do not see clearly? Click for a new code.

User comments