• notice
  • Congratulations on the launch of the Sought Tech site

After walking out of the ivory tower, how does AI really land? Intel gives these two routes

In June 2021, Intel CTO will change coaches. Returning to Intel's CEO Pat Gelsinger again invited his old partner Greg Lavender to serve as CTO.


After taking office, Greg Lavender announced the Developer-first Approach strategy, hoping to provide developers with open and complete hardware, software, tools, technologies and product portfolios with the power of Intel. Greg Lavender said in an interview with InfoQ's "C-bit Face to Face" in November last year that Intel has always paid great attention to the developer community, and also deeply understands the value brought by software developers .


In October last year, at the Intel On Technology Innovation Summit (Intel Innovation), Intel CEO Pat Kissinger reiterated that Intel went back to its roots and re-embraced developers, emphasizing its commitment to the developer community, and Intel's commitment to software and hardware. The developer-first philosophy. At the summit, Intel released the latest products, developer tools and technologies, emphasizing its emphasis on enabling an open ecosystem and ensuring that developers can use the development tools and development environments they favor.


In order to promote the software and hardware collaboration strategy and the developer priority strategy, Intel has made many arrangements in terms of tools, technologies, projects, open source, and ecological communities.

Create a more open source and open software and hardware platform to make AI ubiquitous


Recently, Dai Jinquan, Intel academician and global CTO of big data technology, introduced in an interview with InfoQ and other media that his team mainly focuses on software in AI, big data analysis, machine learning, deep learning and other aspects, from the overall optimization of the underlying database, From the optimization of the framework, to the upper-level AI platform, and to the upper-level, a lot of work has been done for different scenarios, such as recommendation, natural language processing and other aspects of toolsets or algorithm development for a certain field.


"Our goal is to make 'AI everywhere'. On the one hand, Intel's hardware platform can achieve 'everywhere', from small embedded devices, notebooks, to servers, high-performance computing, GPUs. On the other hand, we hope that the entire The hardware plus software platform can expand to 'everywhere'. When AI can run well on Intel hardware, it can help many industries and many users can make better use of AI. We hope that AI It can be used in various places, and AI can be used in different places,” said Dai Jinquan.


The way to achieve this goal is to create an open platform for hardware and software. Among them, open source is a very key focus.


Intel has been actively investing in open source and its ecological construction.


On open source projects such as TensorFlow, PyTorch, MXNet, and XGBoost, the Intel team and the open source community have done a lot of work together, including how to make various optimizations better on Intel's platform, how to make full use of Intel's hardware architecture, how to Let these software itself have a more open architecture and so on.


For example, more and more people want TensorFlow to support different hardware platforms. However, it is not easy to implement this idea in an architecture like TensorFlow, because it is a very large system, with a large user base, and many different hardware platform companies want to support this function.


In order to solve this problem, an architect of Intel proposed and designed a Pluggable device API in the TensorFlow community, which has the function of pluggable device. This is a relatively general architecture, which can not only support Intel hardware, such as Intel's CPU, GPU, can also support various accelerators. The architect also worked with Google's architects to turn the pluggable device functionality into TensorFlow's standard architecture and implement it on top of it. In this way, Intel's hardware platform, as well as many other third-party hardware platforms, can be implemented through this architecture TensorFlow.


In addition, Intel has developed many open source tools, including OpenVINO, BigDL, Analytics Zoo, INC (Intel Neural Compressor), etc., through the development of open source tools, to create an open source community and better serve developers.


Dai Jinquan said that the team is also committed to creating a more open source and open software and hardware platform, on which customers/users can develop their application scenarios more efficiently, and can also apply various technological innovations to their solutions. This will greatly improve the efficiency of AI application development.

AI Practice Day: Aiming at the bottleneck of AI landing


In addition to tools and products, Intel is also working on ecological construction.


Two years ago, Intel launched an "AI Practice Day" in China. This is a project that combines ecology, industry-university-research and talent training. In layman's terms, it is an AI practical course for developers.


Xia Lei, chief engineer of Intel, chief architect of artificial intelligence technology in China, and one of the main initiators of the AI practice day, introduced that the original intention of the AI practice day was very simple and straightforward.


"Intel has developed a variety of innovative technologies and innovative tools, and we hope to deliver these excellent tools and technologies and the latest practical experience to developers through a more effective 'pipeline', and shorten the time from Intel to the market. , The distance to the customer. The AI practice day is such a 'pipeline' role," Xia Lei said.


In the early days, the AI practice day was mainly to pass on the results of Intel's own technology and some open source partners. During the progress of the project, the Intel team found that with the deepening of the interaction with the domestic ecological industry, more innovations and practical results emerged. Intel's ecological partners also have a strong willingness to share their successful experiences and practical results at the AI Practice Day. Therefore, the current AI practice day is not only the home of Intel, but also Intel's partners around the world.


Since its inception, Intel AI Practice Days have covered approximately tens of thousands of developers each year. According to statistics, in 2021, the AI Practice Day will cover 65,000 developers. Moreover, Intel continues to optimize and improve the AI practice day every year, and the iteration is mainly reflected in two aspects:


One is to more accurately segment different customers and different developer audiences. For example, different developers have different needs. Some AI developers focus on algorithm innovation, while others focus on deployment. There is also the division of industry dimensions. Taking the financial industry as an example, developers in the financial industry have not only paid attention to the application of AI in recent years, but also concerned about data privacy, compliance and security issues at the data application level after the popularization of AI. Based on this, Intel customizes the content of AI practice days for different industries and the roles of different developers.


Second, from a technical point of view, this year is the year when Intel is preparing to advance the entire AI product line from CPU to GPU. Therefore, this year's AI practice day hopes to spread the technology of Intel's different application platforms.


Xia Lei believes that at this stage, AI has moved from the ivory tower and the academic circle to industrial practice. However, at this stage, there are still many challenges in AI applications. One of the problems that Intel hopes to solve through the AI Practice Day is to achieve more popular and more popular AI.


At present, many enterprises have the need to apply AI, but due to factors such as talents, capabilities, and experience, they cannot implement AI practices in the end. From the point of view of software tool design, to solve this problem, it is necessary to design software tools to be simpler and easier to use, so as to reduce the threshold for developers to use and the cost of learning. With these tools, easy-to-use solutions can be communicated and interacted with developers more quickly through the AI practice day, thereby accelerating the implementation of AI applications and the popularization of AI applications in various fields.


On the other hand, the AI Practice Day hopes to solve the bottleneck problem faced by AI in the implementation of various industries.


Taking manufacturing as an example, Intel is promoting the concept of device-edge-cloud collaboration. "End" refers to the realization of real-time intelligent manufacturing at the manufacturing site, such as common defect detection. At the production line site, the defect of a certain product should be found immediately, and the defective product should be intercepted from the production line immediately. In the aspect of "edge", the multiple production lines of the factory are managed through the nodes of the edge. In terms of "cloud", the cloud can collect all the data of each factory and each assembly line, and perform in-depth processing and utilization of the data. Intel currently uses OpenVINO to provide better implementation on the edge, and BigDL on the edge and cloud to provide quick deployment and system scalability.


Tags

Technical otaku

Sought technology together

Related Topic

1 Comments

author

buy atorvastatin 10mg generic & lt;a href="https://lipiws.top/"& gt;order atorvastatin pill& lt;/a& gt; cheap lipitor 10mg

Oalrve

2024-03-07

Leave a Reply

+