Liu Qingfeng, deputy to the National People's Congress and chairman of iFLYTEK: Accelerate the construction of autonomous and controllable large models and industrial ecology on domestic computing platforms
This year, China's amazing breakthrough in the performance of large artificial intelligence models has attracted widespread attention in the international field. In the context of the deep reconstruction of the global scientific and technological competition pattern, the training of domestic large models is highly dependent on imported computing power. If we cannot speed up the solution of the weak domestic independent and controllable artificial intelligence industry ecology and difficulties in adaptation, it is tantamount to "building a high-rise on someone else's foundation." On March 4, the reporter learned from iFlytek that Liu Qingfeng, a member of the National People's Congress and chairman of iFlytek, submitted a number of proposals this year. One of the proposals is "Accelerating the Construction of Independent and Controllable Large Models and Industrial Ecology on Domestic Computing Platforms." Liu Qingfeng suggested: First, encourage the development and application of large models based on independent and controllable domestic computing platforms. Give special financial support to enterprises that make domestic computing chips and enterprises that use domestic chips to train large models, give resources to the national public computing power, and accelerate the innovation of large model algorithms based on domestic computing power; encourage central state-owned enterprises to give priority to purchasing full-stack independent and controllable large models developed based on domestic computing platforms; give priority to promoting industry vertical applications based on full-stack independent and controllable large models; second, build a mechanism for full sharing of data resources. Relying on China's extensive AI application scenarios, we will actively promote the application of large models in the industrial field and form a data flywheel, so that China can take the lead in obtaining the dividends of the AI industry. Third, we will provide special support for the construction of an ecosystem based on domestic computing power platforms. We will encourage the development of a large model developer ecosystem and open source community based on an independent and controllable computing power base, and provide special support to accelerate the formation of a domestic large model ecosystem and tool chain, and accelerate the development of my country's independent and controllable artificial intelligence industry ecosystem.