The former chip giant is fully embracing the transformation of artificial intelligence. But before the pressure of the old rival Microsoft, there is NVIDIA's interception, followed by Google and other Internet ups and downs, how does Intel use artificial intelligence to illuminate the future? To answer this question, we must start with the architecture of the CPU and GPU. This is a very complicated and academic issue. Aside from some technical levels, this picture is relatively intuitive and allows us to understand the difference between CPU and GPU. The green in the picture is the calculation unit, the orange-red is the storage unit, and the orange-yellow is the control unit. The GPU uses a large number of computing units and an extremely long pipeline, but only very simple control logic. The CPU has complex control logic and many optimization circuits, but the computing power is only a small part. Therefore, the GPU is better at large-scale concurrent computing, and is suitable for computationally intensive programs with high data parallelism (the same algorithm has different data repetitive calculations). The CPU has low requirements on parallelism and low dependence on data, and is suitable for handling various tasks. When the concept of artificial intelligence came to the public again in 2016, Alphago's popularity will quickly heat up the concept of deep learning. For deep learning that requires a large amount of data parallel computing, the advantages of GPU are directly magnified, but from the full path of artificial intelligence The CPU and GPU must work together. Specific to a complete artificial intelligence application, the previous data acquisition and the final information decision are all good at the CPU, and a large amount of data calculation in the middle is better than the GPU. “For example, for face recognition applications, the acquisition of pre-images and the storage of data are best performed on the CPU. The GPU is then responsible for a large number of parallel recognition calculations. The final recognition results need to be correlated with other information for decision making, or with other sensors and channels. When the information is integrated, the CPU is better at it." Song Jiqiang, dean of the Intel China Research Institute, gave an example of ZOL video interview. However, GPUs are not the only platform for parallel computing. If there is no clear data algorithm and a lot of experimentation is needed to determine which algorithm is more suitable, the GPU will be more efficient and flexible. But once the algorithm is fixed, the GPU's work can be replaced by an acceleration chip optimized for artificial intelligence AI. At present, Intel has two kinds of hardware acceleration chips, FPGA and Nervana. It does not require CPU instruction scheduling, and can be directly optimized for hardware and algorithms, and the execution efficiency is higher. This type of dedicated acceleration chip solves problems such as computational density, memory storage bandwidth, etc., and optimizes performance to the best. "The front-end equipment should consider the size, power consumption, price and other issues. The dedicated acceleration chip is the best choice. The acceleration chip is especially suitable for computer vision applications in the front-end equipment, small-scale deep learning network, so it will appear in the drone. , artificial intelligence camera." Dean Song Jiqiang said. Intel's dedicated acceleration chip reduces the impact of parallel computing, and there are three key acquisitions behind it. Intel buys Altera for $16.7 billion On June 2, 2015, Intel announced the acquisition of programmable logic chip giant Altera for $16.7 billion in cash. This transaction became the largest acquisition in the 49-year history of Intel. CEO Ke Zaiqi lamented the acquisition. The amount of the case is too big." Indeed, Intel’s quarterly revenues have just reached $13.2 billion. It seems that Intel's ruin at that time is still in the right place. Altera's FPGAs have become the key to Intel's development of dedicated acceleration chips after being included in Intel's product line. Intel's vice president of acquisitions, Wendell Brooks, said in an interview at the time that the integrated chip of FPGA and Intel processor package would initially bring about 30%-50% performance improvement compared to traditional processor and FPGA independent components. The final performance improvement will be 2-3 times. Although FPGA can also complete high-performance parallel computing, it is not enough to make up for the lack of GPU, so Intel continues to buy and buy. On August 10, 2016, Intel announced a $400 million acquisition of the deep learning startup Nervana System. Nervana is a company with independent intellectual property rights in the chip field, and its Engine chip can compete with the GPU in terms of power consumption and performance. With the Nervana Acceleration Chip, Intel extends the processor business system to gain more voice in the field of artificial intelligence. On March 13 this year, Intel announced the acquisition of Mobileye for $ 15.3 billion, which is the second largest deal for Intel after the acquisition of Altera for $16.7 billion in 49 years. To know that Intel’s net profit for the entire fiscal year of 2016 was only $10.3 billion, Intel spent almost a year and a half to complete the acquisition. why? Mobileye holds a number of key technologies for autopilot, such as cameras and sensor chips. Even analysts say that it has more than 70% market share in driver assistance and anti-collision systems. The acquisition will further strengthen Intel. The position of technology supplier in the field of autonomous driving. President Song disclosed: "After the acquisition, Mobileye's customers have directly become customers that Intel can support. Intel will be able to provide a complete vehicle solution from algorithm to chip, multi-sensor fusion to storage and communication technology in the field of automatic driving. Driverless technology is not a technology, but a combination of many technologies, Intel will be very good at creating an efficient and reliable driverless platform." In the past three decades, from PC to mobile, human-computer interaction technology has taken a large proportion. In the era of artificial intelligence, the machine will have a better subjective initiative, and the task will be completed by the sensor data autonomously aware task, and no one needs to initiate it. However, Dean Song believes that artificial intelligence technology still has many limitations, and it must be combined with human-computer interaction to better protect the user experience. “Artificial intelligence can train a good model and achieve 99% reliability, but in the application process, it may encounter new situations that have not been touched before, and the accuracy of artificial intelligence may not be as high. Intelligence and human-computer interaction should be combined, even if artificial intelligence is wrong, it will not cause errors. Artificial intelligence can further reduce the time for people to actively operate, but still can interact with people to confirm the best results." The voice assistant represented by Siri was advertised as an early artificial intelligence model, but many users felt that they were not useful. In the view of Dean Song, it is mainly limited by the form of equipment, and the space for artificial intelligence is not large. "Many operations on mobile phones are not complicated. People's operating costs are not high and they must be more accurate than machine operations, so you won't feel a revolutionary, disruptive experience. But if the product form is not a mobile phone, but a smart car, The value brought by drones and intelligent robots is unprecedented.†In the era of intelligent robots, you can buy a good intelligent robot with the price of a high-end smartphone. It will make user portraits for all family members. Provide differentiated and customized services for different people's characteristics and needs. If intelligent robots have mature space movement and autonomous operation capabilities, intelligent pensions have hopes to achieve. Today, Intel's artificial intelligence solutions have quietly appeared in many application scenarios. In 2016, the double 11 was in the peak period of Jingdong users' access. Intel's artificial intelligence technology monitors the uploading of illegal images in real time, guarantees the operational safety of Jingdong traffic during peak periods, and helps Jingdong analyze customer trends in detail and provide more accurate recommendations. In the Xunfei Hearing-Intelligent Conference System, all voices can be converted into text in real time with an accuracy rate of over 95%. In addition to the core voice technology of Keda Xunfei, Intel's Xeon processor and real-time voice transfer Write a solution; by 2020, it takes only 24 hours to complete a series of precise medical procedures, from gene sequencing to medical analysis to medical planning. The Intel GTX One Biocomputing Acceleration Platform is designed to rely on the computing power of artificial intelligence to drive precision. Medical rapid development. . . Next, we will soon see Intel AI's technologies and solutions released in the virtual/augmented reality and driverless areas. In the VR/AR world, in addition to the user's ability to interact with the virtual environment using his own hands, Intel's artificial intelligence technology will enable the user to wear a helmet, and people who interact with him can still see his face. Intel's driverless technology will provide a smart interactive experience in the cockpit, and the driver can let go of his hand to do other things, but artificial intelligence will monitor you, prevent you from being distracted, and can pull you in an emergency to do the corresponding task takeover. . At one time, Intel's first microprocessor became an important milestone in the history of computers. However, when the mobile Internet gradually obscured the PC's light, Intel did not get out of the fog, and the loss of the mobile bonus gave the Intel strong man the determination to cut the wrist. Today, Intel has tore off the label of the "chip giant" and replaced it with "drive everything smart interconnect." Throughout Intel's key deployment areas in 2017, whether it is artificial intelligence, driverless, or virtual reality, 5G, Intel wants to fully attack the hope to turn the tide, but this time, it is not too late. Membrane Switch Assemblies,Lcd Display Membrane Switch,Tactile Membrane Switch Keypad,Overlay Tactile Membrane Switch KEDA MEMBRANE TECHNOLOGY CO., LTD , https://www.kedamembrane.com