Artificial intelligence (AI) is the core of the development of Internet of Things and Industry 4.0. In particular, when Tesla launched the electric car and Apple (Apple) released the new machine iPhoneX launched FaceID, let the market experience to the AI chip unlimited business opportunities. At the same time, AI application of the higher acceptance of the country, will contribute to its GDP greater.
AI chips include three major categories of markets, namely data center (cloud), communication terminal products (mobile phones), specific application products (self-driving, put on AR / VR, UAV, robot ...). The current machine learning and more use of GPU image processing, especially Nvidia is a leader in this field, but some industry that GPU processing efficiency is not fast enough, and in response to the different needs of many new products, so, the introduction of NPU, VPU, TPU, NVPU ... and many more. It is unclear which architecture of the chip will win in the AI war. But the (mobile) terminal market for the AI chip power consumption, size, price has a very strict requirements, the difficulty is higher than the cloud data chip. To grab the future AI application market opportunities, technology giants such as Google, Microsoft, Apple's attempt to build AI platform eco-mode to eat the entire industry chain.
Figure: Artificial intelligence is the core of the Internet of Things
At present, the future development of AI has eight new trends:
一、AI in the vertical application of the industry has great potential
The artificial intelligence market has great potential in the vertical areas of retail, transportation and automation, manufacturing and agriculture. The main factor driving the market is the increasing number of applications of artificial intelligence technology in the vertical field of various end users, especially improving the end consumer service.
Of course, the artificial intelligence market is also up to the IT infrastructure is also perfect, smart phones and intelligent wear-wearing equipment. Among them, the natural language processing (NLP) application market accounted for a large part of the AI market. With the natural language processing technology continues to sophisticated and drive the growth of consumer services, as well as: automotive information and communication entertainment systems, AI robots and AI-enabled smart phones and other fields.
二、AI into the health care industry to maintain high-speed growth
As the health care industry uses large amounts of data and artificial intelligence, and thus improve the diagnosis of disease, medical personnel and patients between the human imbalance, reduce medical costs and promote cross-industry cooperation. In addition, AI is also widely used in clinical trials, large medical programs, medical advice and promotion and sales development. AI into the health care industry from 2016 to 2022 to maintain a high growth, is expected from 2016 to 667.1 million US dollars in 2022 to $ 79.888 billion annual average compound growth rate of 52.68%.
三、AI to replace the screen to become a new UI / UX interface
In the past from the PC to the mobile phone era, the user interface is through the screen or keyboard to interact. With smart speakers (SmartSpeaker), virtual / enhanced reality (VR / AR) and autopilot system into the human environment, to accelerate in the absence of the screen, people can easily and freely communicate with the computing system. This means that artificial intelligence through the natural language processing and machine learning to make technology more intuitive, but also become easier to control, the future will be able to replace the screen in the user interface and user experience status. In addition to playing an important role in the back end of the enterprise, artificial intelligence can take on more complex roles in the technical interface. For example: the use of visual graphics of the automatic driving car, through the artificial neural network to achieve real-time translation, that is, artificial intelligence makes the interface easier and more intelligent, and therefore set the future of high standards of interaction mode.
四、the future mobile phone chip must be built AI core
At this stage the mainstream of the ARM architecture processor speed is not fast enough, to carry out a large number of image operations are still inadequate, so the future of mobile phone chip will be built AI core. As, Apple will be 3D sensing technology into the iPhone, Android camp smartphone will be in next year (2017) follow-up into 3D-related applications.
五、AI chip is the key to successful integration of hardware and software
The core of the AI chip is semiconductors and algorithms. AI hardware is mainly required faster instruction cycle and low power consumption, including GPU, DSP, ASIC, FPGA and neuron chip, and must be combined with the depth of learning algorithms, and the key to success is the advanced packaging technology. Overall, the GPU is faster than the FPGA, and the power efficiency is better than the GPU, so the AI hardware option depends on the needs of the product supplier. For example, Apple's FaceID face recognition is the 3D depth sensor chip plus the neural engine computing function, integrated up to 8 components for analysis, namely infrared lens, flood light sensor components, distance sensors, ambient light sensor, front camera, point Array projector, speaker and microphone. Apple emphasizes the user's biometric data, including: fingerprint or face recognition are stored in the form of encrypted inside the iPhone, so easy to be stolen.
六、AI self-study is the ultimate goal
AI "brain" smart is a phased, from the machine learning evolution to the depth of learning, and then evolved to independent learning. At present, still in the machine learning and in-depth learning stage, to achieve independent learning needs to solve the four key issues. First, it is an autonomous platform to create an AI platform; but also to provide a self-learning machine can make independent learning virtual environment, must comply with physical laws, collision, pressure, the effect should be the same with the real world; and then AI " "Into the framework of the autonomous machine; the final establishment of the virtual world entrance (VR). At present, NVIDIA introduced its own machine processor Xavier, in preparation for the business and popularization of autonomous machines.
七、the most perfect architecture is the CPU and GPU (or other processors) together
The future, but also introduced a number of specialized areas required super performance of the processor, but the CPU is common in a variety of equipment, what the scene can be applied. So, Harvatek the most perfect architecture is the CPU and GPU (or other processors) together. For example, NVIDIA introduced the CUDA computing architecture, which combines dedicated ASICs with common programming models to enable developers to implement multiple algorithms.
八、AR become AI's eyes, the two are complementary, indispensable
The future of the AI need AR, the future of the AR also need AI, AR can be compared to AI's eyes. For the robot to create and create in the virtual world, itself is a virtual reality. Also, if you want to enter the virtual environment to the robot training, but also need more other technology. As to whether the CPU will be TPU, NPU, VPU ... and so on to replace the new type of processor, the answer should not. Because the emerging processor is only to deal with new or unresolved issues, and the future tend to CPU integration. At the same time, the chip market expectations to have more competition and choice, not Intel, Qualcomm dominate. To meet the Internet of Things era, in the past we believe that Moore's Law will eventually reach the limit, but the future of silicon is heterogeneous and cross-border integration, there are many needs did not appear. NVIDIA CEO Huang Renxun said that Moore's Law is already the old rule, GPU computing rate and neural network complexity in the past 2 to 5 years showed explosive growth.
Looking ahead, with AI, Internet of Things, VR / AR, 5G and other technologies mature, will drive a new wave of semiconductor industry 30 years of prosperity, including: memory, central processing unit, communication and sensor four chips, Product application chip demand continues to increase, to China's huge market advantage in the semiconductor industry can play a key role in the world.
The artificial intelligence market has great potential in the vertical areas of retail, transportation and automation, manufacturing and agriculture. The main factor driving the market is the increasing number of applications of artificial intelligence technology in the vertical field of various end users, especially improving the end consumer service.
Of course, the artificial intelligence market is also up to the IT infrastructure is also perfect, smart phones and intelligent wear-wearing equipment. Among them, the natural language processing (NLP) application market accounted for a large part of the AI market. With the natural language processing technology continues to sophisticated and drive the growth of consumer services, as well as: automotive information and communication entertainment systems, AI robots and AI-enabled smart phones and other fields.
二、AI into the health care industry to maintain high-speed growth
As the health care industry uses large amounts of data and artificial intelligence, and thus improve the diagnosis of disease, medical personnel and patients between the human imbalance, reduce medical costs and promote cross-industry cooperation. In addition, AI is also widely used in clinical trials, large medical programs, medical advice and promotion and sales development. AI into the health care industry from 2016 to 2022 to maintain a high growth, is expected from 2016 to 667.1 million US dollars in 2022 to $ 79.888 billion annual average compound growth rate of 52.68%.
三、AI to replace the screen to become a new UI / UX interface
In the past from the PC to the mobile phone era, the user interface is through the screen or keyboard to interact. With smart speakers (SmartSpeaker), virtual / enhanced reality (VR / AR) and autopilot system into the human environment, to accelerate in the absence of the screen, people can easily and freely communicate with the computing system. This means that artificial intelligence through the natural language processing and machine learning to make technology more intuitive, but also become easier to control, the future will be able to replace the screen in the user interface and user experience status. In addition to playing an important role in the back end of the enterprise, artificial intelligence can take on more complex roles in the technical interface. For example: the use of visual graphics of the automatic driving car, through the artificial neural network to achieve real-time translation, that is, artificial intelligence makes the interface easier and more intelligent, and therefore set the future of high standards of interaction mode.
四、the future mobile phone chip must be built AI core
At this stage the mainstream of the ARM architecture processor speed is not fast enough, to carry out a large number of image operations are still inadequate, so the future of mobile phone chip will be built AI core. As, Apple will be 3D sensing technology into the iPhone, Android camp smartphone will be in next year (2017) follow-up into 3D-related applications.
五、AI chip is the key to successful integration of hardware and software
The core of the AI chip is semiconductors and algorithms. AI hardware is mainly required faster instruction cycle and low power consumption, including GPU, DSP, ASIC, FPGA and neuron chip, and must be combined with the depth of learning algorithms, and the key to success is the advanced packaging technology. Overall, the GPU is faster than the FPGA, and the power efficiency is better than the GPU, so the AI hardware option depends on the needs of the product supplier. For example, Apple's FaceID face recognition is the 3D depth sensor chip plus the neural engine computing function, integrated up to 8 components for analysis, namely infrared lens, flood light sensor components, distance sensors, ambient light sensor, front camera, point Array projector, speaker and microphone. Apple emphasizes the user's biometric data, including: fingerprint or face recognition are stored in the form of encrypted inside the iPhone, so easy to be stolen.
六、AI self-study is the ultimate goal
AI "brain" smart is a phased, from the machine learning evolution to the depth of learning, and then evolved to independent learning. At present, still in the machine learning and in-depth learning stage, to achieve independent learning needs to solve the four key issues. First, it is an autonomous platform to create an AI platform; but also to provide a self-learning machine can make independent learning virtual environment, must comply with physical laws, collision, pressure, the effect should be the same with the real world; and then AI " "Into the framework of the autonomous machine; the final establishment of the virtual world entrance (VR). At present, NVIDIA introduced its own machine processor Xavier, in preparation for the business and popularization of autonomous machines.
七、the most perfect architecture is the CPU and GPU (or other processors) together
The future, but also introduced a number of specialized areas required super performance of the processor, but the CPU is common in a variety of equipment, what the scene can be applied. So, Harvatek the most perfect architecture is the CPU and GPU (or other processors) together. For example, NVIDIA introduced the CUDA computing architecture, which combines dedicated ASICs with common programming models to enable developers to implement multiple algorithms.
八、AR become AI's eyes, the two are complementary, indispensable
The future of the AI need AR, the future of the AR also need AI, AR can be compared to AI's eyes. For the robot to create and create in the virtual world, itself is a virtual reality. Also, if you want to enter the virtual environment to the robot training, but also need more other technology. As to whether the CPU will be TPU, NPU, VPU ... and so on to replace the new type of processor, the answer should not. Because the emerging processor is only to deal with new or unresolved issues, and the future tend to CPU integration. At the same time, the chip market expectations to have more competition and choice, not Intel, Qualcomm dominate. To meet the Internet of Things era, in the past we believe that Moore's Law will eventually reach the limit, but the future of silicon is heterogeneous and cross-border integration, there are many needs did not appear. NVIDIA CEO Huang Renxun said that Moore's Law is already the old rule, GPU computing rate and neural network complexity in the past 2 to 5 years showed explosive growth.
Looking ahead, with AI, Internet of Things, VR / AR, 5G and other technologies mature, will drive a new wave of semiconductor industry 30 years of prosperity, including: memory, central processing unit, communication and sensor four chips, Product application chip demand continues to increase, to China's huge market advantage in the semiconductor industry can play a key role in the world.
