[The upcoming 5G brings artificial intelligence closer and closer to us? ] In 2007, people used to call it the beginning of the mobile era, because that year, the Internet broke out, social networks emerged, Google, Microsoft, and Facebook entered the public’s perspective. Two things changed the world: One is Apple. The first iPhone was released at the beginning of the year, and the second was Qualcomm’s first-generation Xiaolong chip in November of the same year. In 2007, Qualcomm Research began its first personal-intelligence project to study pulsed neural networks for computer vision and motion control applications. At that time, Qualcomm might not have expected it. Ten years later, it will pioneer mobile communications and artificial intelligence. The direction is becoming the biggest force for change in the digital world today.
With regard to artificial intelligence, we have heard many miraculous stories. For example, AlphaGo has defeated humans under Go, but under the impetus of these two revolutionary forces, technology is subtly changing the mobile terminal. You may have noticed that the mobile phone camera is more beautiful, the sweeping robot is working smarter, the smart speaker knows you better... More importantly, this revolutionary era has only just begun because of the upcoming 5G, The terminals around us can carry a greater amount of computing power.
Walking on the "edge" of intelligence
Just as Meng Pu, chairman of Qualcomm China, recently mentioned at the Qualcomm Artificial Intelligence Innovation Forum: “Artificial intelligence is affecting more and more industries and gradually expanding from the cloud to the terminal, changing people’s lives and working methods.â€
In the past, many trainings on neural networks were done in the cloud or on servers. In recent years, the entire model has undergone many changes. Some artificial intelligence training, execution, and inference work, such as pattern matching, modeling detection, classification, identification, detection, etc., have gradually shifted from the cloud to the terminal side, and all edge terminals Both will have machine learning capabilities. This means that data processing will be processed at the location closest to the data source, supplementing the cloud processing.
There are three main reasons driving this migration and evolution:
First, data privacy. Consumers value personal data privacy very seriously. The sharing of personal data should be better protected. Terminal-side security will be better.
Second, performance. Completion of related processing and inference work on the terminal side can maintain ultra-low delay operation and operation. If everything is completed in the cloud, taking into account the round-trip time for cloud processing to obtain information, there will undoubtedly be a great deal of delay. challenge.
Third, reliability. The terminal cannot completely rely on the cloud to complete all AI operations and processing. Once the connection is broken, the terminal will lose its own value.
The value of terminal-side artificial intelligence is beyond doubt, and Qualcomm, as a mobile communications company, has an advantage in the terminal side. Gary Brotman, director of product management at Qualcomm, once mentioned in an interview that Qualcomm, as a hardware platform and computing platform provider, is to ensure that the terminal-side AI can run efficiently from performance to endurance and achieve a better user experience.
Focus on smart and 5G Qualcomm
Wireless communication technology has seen a leap every decade, and Qualcomm has been hoping to change the way the world connects, calculates, and communicates for more than 30 years. Recalling the history of Qualcomm, we first realized the digitization of mobile communications and ensured a mobile phone. Second, through the smart phone, the calculation was redefined. Now, Qualcomm is very excited about the upcoming transformation of 5G and artificial intelligence. Qualcomm president Cristiano Amon used the term “expectation†more than once during the Qualcomm artificial intelligence innovation forum. This change.
Qualcomm president Cristiano Amon
According to data from Gartner in March 2018, by 2021, the commercial value derived from artificial intelligence will reach 3.3 trillion US dollars.
Amon believes that with trillions of terminals connected to each other, to achieve scale, intelligence must be distributed to the wireless edge. Of course, this is not an easy task. Once machine learning is done at the edge, it means that you need to complete multiple types of tasks at the same time in a limited environment. Therefore, to truly implement artificial intelligence at the edge, you need to provide different solutions.
5G is an important solution. The ultra-high-speed, low-latency 5G network supports rapid connection to the cloud, and access to unlimited storage and data in the cloud. At the same time, it senses, infers, and acts on endpoints with processing capabilities. In other words, the future 5G network, in addition to providing unlimited data access and high-speed connectivity with the cloud, will also bring about a huge increase in wireless edge computing capabilities. For example, this combination of AI and 5G can enable smart devices to achieve the same performance as PCs, such as enjoying the same VR experience as a dedicated gaming PC on a mobile phone.
Even more exciting is that from 2018 to 2022, cumulative shipments of smart phones will exceed 8.6 billion, and the scale of mobile terminals will bring great potential for artificial intelligence platforms. Undoubtedly, the huge scale of smart phones plus Shanghai's IoT will bring artificial intelligence to trillions of connected terminals.
This is why Qualcomm chose to expand intelligence to the terminal side, transform artificial intelligence, use its 5G advantages to build processing capabilities at the wireless edge, and enhance application-oriented computing performance.
As mentioned at the beginning of the article, Qualcomm began research on artificial intelligence many years ago. At present, research has been applied to the product side - from the first generation of artificial intelligence platform Xiaolong 820, to the third generation platform Xiaolong 845, Qualcomm mobile platform has been prepared for the 5G era. QualcommAIResearch was established at the Qualcomm Artificial Intelligence Innovation Forum to focus on the research of artificial intelligence under a unified framework to accelerate the innovation of Qualcomm's artificial intelligence on the terminal side.
It is worth mentioning that terminal artificial intelligence also has challenges.
It is not difficult to find that mobile terminals are driving the continuous evolution of engineering design. In the 4G era, we need to provide mobile devices with the same performance as PCs while considering the design of the terminal. The future will also be like this, requiring artificial intelligence to handle more concurrent applications and more machine learning scenarios. At the same time, it should also take into account efficient cooling and shape design; consumers will also expect more battery life than the whole day, and in the future there may be “unlimited storageâ€. This requires ensuring high-speed data transmission while guaranteeing connectivity; and in terms of terminal-side artificial intelligence workloads, its challenges include supporting intensive calculations, continuous processing of large and complex models, and having complicated concurrency. Sex, because these processes are done on smartphones, so it will also be always on and real time.
These challenges put forward new requirements for terminal design. At the same time, they have to think about chip manufacturers and OEMs. In the 5G era, how smart phones should implement artificial intelligence.
In this regard, Amon stressed that Qualcomm is committed to bringing ubiquitous smart experiences to life, focusing mainly on two aspects: high-efficiency artificial intelligence and personalized artificial intelligence.
Artificial intelligence will provide all relevant information about the user in the smart phone, provide relevant information in the appropriate environment, or meet the needs of the user - this requires very efficient hardware, algorithmic improvements, software tools, and system-level solutions. This is one of the key attributes of the Dragon Platform.
A few months ago, Qualcomm released the Xiaolong artificial intelligence engine (AIE, AIEngine). In just a few months, AIEngine is integrated on different platforms. Now, Qualcomm continues to expand the artificial intelligence ecosystem.
In terms of products, Qualcomm launched the Xiaolong 710 mobile platform, adopts an efficient architecture that supports artificial intelligence, integrates AIEngine, and has neural network processing capabilities. Kedar Kondap, vice president of product management for Qualcomm, said: “The Xiaolong 710 is designed to transform our customers' products into the ultimate personal assistants and enhance the key consumer's daily experience. For example, high-end camera features will benefit from high-speed AI processing at the terminal side without the need for Sacrifice battery life."
In terms of cooperation, Qualcomm launched a AI development kit called TurboXAIDeveloperKit with Thundercomm in Chongqing to provide developers with available artificial intelligence engines, artificial intelligence reference applications and models such as object recognition, defect detection, and scene detection. Pet identification, it will also use modular design, support for the expansion of AI and shooting capabilities; Qualcomm and Baidu cooperation, using AIEngine, through the ONNX (OpenNeuralNetworkExchange) exchange format, to promote Baidu PaddlePaddle open source deep learning framework model in Qualcomm? The conversion and application of the mobile platform; In addition, Qualcomm and NetEase cooperation, using AIEngine components, accelerate the realization of real AR rendering in some part of the dragon platform, which is the first time that the new features are implemented on the Android platform. In other words, as long as you open the Daoist Translation Officer and aim the mobile phone camera at the text content to be translated, real-time AR translation of Chinese, English, Japanese, and Korean can be achieved without taking pictures or relying on the Internet or the cloud for processing.
Amon said, “We hope to provide such computing power for every terminal equipped with a platform.†In the future, Qualcomm will work with a number of operators to build computing power on the edge of the 5G network and use 5G performance and low. Delays further release the potential for edge calculations. At the same time, Amon said frankly that if there is no extensive industry cooperation, Qualcomm will not be able to do all this.
Pin Header Connector,Pin Header Female,Male Header Pins,Right Angle Pin Header
Cixi Xinke Electronic Technology Co., Ltd. , https://www.cxxinke.com