Beijing plans to supply the numeracy needed for AI training as a whole and integrate the large model Chinese corpus.
CTOnews.com, May 17, Beijing issued a public announcement on May 12 on soliciting opinions on several measures of Beijing to promote the Innovation and Development of General artificial Intelligence (2023-2025) (draft for soliciting opinions) (hereinafter referred to as "draft for soliciting opinions"), with a view to implementing the overall supply of computing power needed for AI training.
The draft for soliciting opinions proposes to strengthen the overall supply capacity of computing resources, strengthen cooperation with market players such as head public cloud manufacturers, implement the computing partnership program, and determine the first batch of partner program members. clearly define the supply technical standards, software and hardware service requirements, computing supply scale, preferential strategies, etc., and announce a number of high-quality numerate suppliers to colleges and universities in Beijing and small and medium-sized enterprises.
According to the consultation draft, the government's unified import will be used to reduce the procurement costs of public clouds, benefit small and medium-sized enterprises, and reduce the communication costs of enterprises facing different cloud manufacturers. To meet the demand of flexible computing power, build a unified multi-cloud computing power scheduling platform to achieve unified management and unified operation of heterogeneous computing environment, and facilitate enterprises to run all kinds of artificial intelligence computing tasks seamlessly, economically and efficiently in different cloud environments. Build a directly connected basic optical transmission network between Beijing and Hebei, Tianjin, Shanxi, Inner Mongolia and other provinces (cities), further enhance the platform's ability to perceive the computing resources of the four places, and explore to carry out arithmetic transactions.
The draft also said that in view of the fact that the proportion of high-quality Chinese corpus for large model training is too small, which is not conducive to Chinese contextual expression and industrial application, integrate existing open source Chinese pre-training data sets and high-quality Internet Chinese data and conduct compliance cleaning. At the same time, we will continue to expand high-quality multimodal data sources, build compliant and safe Chinese, picture-text pairs, audio, video and other large model pre-training corpora, and open them conditionally through the social data area of Beijing International big data Exchange.
CTOnews.com attached "some measures to promote the Innovation and Development of General artificial Intelligence in Beijing (2023-2025) (draft for soliciting opinions)" complete document: click here to view
