Most popular at the first world science and techno

At the first World Forum on science and technology and development, experts pointed out that open source is one of the main trends in the development of artificial intelligence

technology Zhang Galen

"flying oar", as the name suggests, wants to let developers in the field of deep learning cross the river and dock smoothly. On October 16, at the first World Forum on science and technology and development, baidu released the open source open platform for in-depth learning at the industry level of flying oars

based on Baidu's years of in-depth learning technology, research and business application of the 10000 ton carbon graphite anode material invested by chengxinneng, and in-depth cooperation with Huawei and other chip enterprises, according to the data of the U.S. Geological Survey, it integrates the core framework of in-depth learning, basic model library, end-to-end development kit, tool components and service platform, and was officially open source in 2016. Its open source has lowered the threshold of AI technology and has served more than 1.5 million developers

in today's era, "open source" is a word constantly mentioned. Wang Haifeng, chief technology officer of Baidu, said in an interview with science and technology that open source is one of the global trends in the development of artificial intelligence

open source, but also self-control

"deep learning is the core support of the new generation of artificial intelligence." Wang Haifeng introduced that in the technical system of artificial intelligence, the deep learning framework is between the hardware layer and the application layer. Its role is equivalent to windows in the personal computer era and android/ios in the mobile era. It can be called the operating system in the intelligent era. The deep learning platform with the deep learning framework as the core is the key infrastructure for the R & D and industrialization of artificial intelligence technology

however, strong technical support is needed to independently develop a deep learning framework and build a complete deep learning platform. So far, there are only a few enterprises worldwide that can make a fully functional deep learning platform, such as Google, Facebook and so on. And the deep learning open source platform launched by them does occupy a large share in the field of global artificial intelligence, which is also used by many domestic enterprises

however, domestic enterprises can't just think of "leaning against a big tree to enjoy the cool". After all, the in-depth learning framework is so important that if we only rely on a few, there will inevitably be constraints; Moreover, the development characteristics of artificial intelligence in different countries are different, and developers also need a platform that can better meet their own needs

moreover, if the open source deep learning platform is combined with artificial intelligence chips, it can give play to the technical advantages of software and hardware integration optimization, build a "chip + operating system" alliance in the intelligent era, and have the right to speak. In addition, in-depth learning related talents are scarce and the training cost is high. If there is no self built ecosystem, it may be difficult to continuously cultivate in-depth learning related core talents

Wang Haifeng said that the research and development of technologically advanced, fully functional, autonomous and controllable in-depth learning frameworks and platforms plays a vital role in promoting the technological innovation, industrial development and talent training of artificial intelligence in China, realizing the autonomous control of artificial intelligence technology and industry, and accelerating the intelligent upgrading of all walks of life

it is reported that the propeller has been equivalent to foreign frameworks in terms of basic functions, performance, completeness of chip support and other technical indicators, and is also ahead of foreign frameworks in terms of algorithm model of Chinese natural language processing, large-scale distributed training, reasoning engine and other technologies

"although there are still shortcomings in user scale, community activity and ease of use, by giving full play to China's advantages in the application of data and artificial intelligence industry, the domestic deep learning framework can fully compete with foreign countries and be able to lead." Wang Haifeng said

the key to building an open source community is to establish an open source thinking

as of August this year, the open source community GitHub has gathered more than 40 million developers, 2.1 million companies and institutions, and has more than 100million code bases. "Among the popular open source projects on GitHub, artificial intelligence accounts for a large proportion, with an exhibition area of 250000 square meters. Open source is one of the major trends in the development of artificial intelligence." Wang Haifeng said that the rapid development of artificial intelligence has promoted the prosperity of open source and open ecosystem; Open source also promotes the rapid popularization of artificial intelligence, and the two promote each other

github is a model of an open source community. So, how to build a truly active open source community and make up for the shortcomings mentioned above

Wang Haifeng believes that the open source community should be dynamic, and the core is that the community is valuable for developers to strictly follow the steps specified in the instrument instruction manual; It should be able to provide a large number of useful projects, tools and models, and create value for the industry. In this way, developers will spontaneously and actively contribute their own efforts to keep the community prosperous

"the key to building an open source community is to establish the open source thinking of China's software industry." Wang Haifeng suggested that enterprises, universities and scientific research institutions can be encouraged to establish an evaluation and incentive mechanism for open source technology talents to promote the construction of high-end talent teams; Support relevant enterprises to cooperate with universities and scientific research institutes to carry out teacher training on open source technology and application; In the course education of the University, we also cultivate students' awareness of open source, open domestic software related courses, and cultivate more basic software talents

he also called for strengthening the promotion and application of China's independent and controllable open source software and platforms, and encouraging developers to innovate based on China's independent and controllable open source software, platforms and communities through policy guidance, setting up special projects, etc. "In major science and technology and industrialization projects supported by the state, China's independent open-source software can also be used to carry out scientific research, so as to ensure the autonomy, controllability and information security at the bottom." Wang Haifeng said