Not long ago, the "2016-2045 Emerging Technology Trend Report" released by the United States regards artificial intelligence as one of the most noteworthy technological development trends. With the rapid development of artificial intelligence, the problem of insufficient computing power of traditional computer chips has become increasingly prominent. It is imperative to develop a new generation of computer chips that can meet the needs of artificial intelligence computing. The artificial intelligence with the "core" of 澎湃 will greatly enhance the autonomy of robots, drones and other equipment, and realize the tremendous improvement of unmanned combat and cloud computing capabilities. In the future, the development of mature artificial intelligence chips will have powerful computing power, or will revolutionize the computing system.
Traditional processors have been "old clocks"
For a long time, Moore's Law has been driving the rapid development of the computer chip industry as a basic principle. However, with the challenges of the silicon-based limit in the field of precision manufacturing in humans, it has become increasingly difficult to double the number of transistors on an integrated circuit about every 18-24 months. The computer system based on the "von Neumann architecture" has not changed substantially for a long time. The computing power of massive data consumption is enough to make traditional computer chips "stressful."
With the rapid development of artificial intelligence, the requirements for chips are getting higher and higher. The deep learning mode adopted by artificial intelligence is essentially a multi-level artificial neural network algorithm. This kind of algorithm mainly sums up the relevant laws from the input massive data, which requires a large amount of calculation processing for massive data. At this point, the traditional processor architecture requires hundreds or even thousands of instructions to complete the "neuron" reaction process of the human brain, which appears to be "old-aged." Deep learning mode, especially the need to speed up the calculation process.
The "initial heart" of artificial intelligence chip design is to accelerate the deep learning algorithm, to better simulate the neural characteristics of the human brain from the underlying architecture, and to achieve more intelligent calculation. This artificial intelligence chip will better meet the needs of deep learning systems for data calculation, and at the same time adjust the massive parameters, which will become the "catalyst" for the development of artificial intelligence. In this regard, researchers have been seeking to develop a new generation of chips to break through the "bottleneck" of Moore's Law failure, to meet the computational needs of artificial intelligence.
Mining smart chips "infinite possibilities"
At present, the development of artificial intelligence chips is still in the stage of research and development. From the technical architecture point of view, artificial intelligence chips can be mainly divided into general-purpose chips (GPU, FPGA), FPGA-based semi-customized chips, fully customized ASIC chips for deep learning algorithms, and brain-like computing chips. The typical representative is the TPU chip of Google Inc. and the "Cambrian" chip developed by China.
In the man-machine Go game, Google "Alpha Dog" uses about 170 GPUs and 1200 CPUs, and it occupies a computer room and is equipped with high-power air conditioners. If you replace it with an artificial intelligence chip, you can replace them with a box-sized space. At present, the prototype of brain-like computational artificial intelligence chips has emerged. IBM's Truenorh chip contains 1 million digital neuron arrays and 256 million communication synapses, which can basically simulate the data processing of human brain neurons. At the same time, the new concept chip represented by pulse neural network chip DeepSouth and deep learning brain neuron chip DeepWell also indicates the future development direction of artificial intelligence chip.
The US Department of Defense and the Air Force Research Laboratory have shown great interest in the outstanding performance of TPU chips in the field of machine learning, and have long begun to explore the application of artificial intelligence technology in the military field. At present, they are conducting discussions with Google to further enhance the combat capability and intelligence level of US military unmanned systems. In the future, the US military will also explore the use of TPU chips to deploy "Military Cloud 2.0" to enhance the intelligent deployment of UAVs and unmanned systems.
The field of network security is showing great power
With the continuous advancement of the new generation of artificial intelligence chip projects of the US Department of Defense Advanced Research Projects Agency, artificial intelligence chips will be in the field of network security. The artificial intelligence chip-based autonomous network attack system developed by Stanford University in the United States can independently learn the network environment and generate specific malicious code to achieve attacks on the specified network and information theft. This kind of autonomous network attack system with artificial intelligence has strong self-learning ability and virus defense capability, and has been highly valued and invested by the US Defense Advanced Research Projects Agency.
Once the artificial intelligence autonomous network attack system is successfully developed, it can further enhance the offensive strength of the US military in the field of network security. The new autonomous network attack system is mainly based on the general architecture of the deep neural network processor, and only the basic autonomous learning system program is built in. When it runs in a specific network, it can learn the architecture, scale, and device type of the network autonomously. By intercepting and analyzing the network data, it can independently write a set of attack code suitable for the network environment every 24 hours. The attack program is dynamically adjusted in real time according to the network environment. Since the attack code is basically "freshly baked", the existing antivirus system relying on the virus database and behavior recognition is difficult to identify, and thus the system is highly concealed and destructive.
In addition to the artificial intelligence autonomous network attack system, the US Department of Defense Advanced Research Projects Agency has added a research project such as “cerebral cortex processor†as early as 2015, aiming to develop better data processing by simulating the structure of human cerebral cortex. New brain-like chips. These artificial intelligence chips have extremely low power consumption and can be used for real-time data-aware processing and target recognition. They can also solve the problems of real-time control of high-speed moving objects. After the application is put into use, the unmanned combat and cloud computing capabilities will be greatly improved. .
It is foreseeable that people will also find network vulnerabilities through artificial intelligence to further enhance the network attack capability, and the future network security will face more severe challenges. 
Storage Electric Water Heater,Electric Water Heates For Showers,Enamel Inner Tank Electric Water Heater,Wall Mounted Storage Water Heater
Shandong Sangle Group Co.,Ltd. , https://www.sangle-group.com