China Telecom’s Breakthrough: Training a Trillion-Parameter AI on Domestic Chips

AI News

2 Mins Read

In-Short

  • China Telecom develops two ‍LLMs using domestically-produced chips.
  • Breakthrough⁢ signifies China’s push for AI self-reliance amid US semiconductor restrictions.
  • One‌ model boasts one trillion parameters,⁤ rivaling⁢ mainstream AI platforms.
  • Domestic chipmakers like Huawei’s Ascend chips are central​ to China’s AI strategy.

Summary of China ⁤Telecom’s AI⁢ Advancements

China Telecom, ​a⁣ major state-owned telecom ⁢entity in China, has made significant ‍strides in artificial intelligence ⁣(AI) by ‌training⁢ two ‍large language models ⁢(LLMs)​ on chips produced ‌within the country. This development​ is‌ a ⁤direct‌ response to the United States’ stringent export controls ‌on advanced‍ semiconductors, which have impacted Chinese companies’ ​access to high-end processors ​from firms like ⁣Nvidia.

The AI institute of China Telecom announced the success ⁢of training ‍the models,⁣ TeleChat2-115B and an unnamed model with one trillion​ parameters, on thousands of ‌Chinese-made chips. This achievement is a testament to China’s growing⁤ capabilities in independently developing AI technology, marking a shift towards innovation and self-sufficiency.

Chinese companies are⁣ accelerating their ​efforts⁤ to match the AI prowess of global leaders, especially ‍as US export⁤ restrictions limit their access to⁢ cutting-edge AI chips. By creating their own processors, such as the ⁢TeleChat2-115B with roughly 100⁣ billion parameters, China is reducing its reliance on Western technology.

While China Telecom has not disclosed‌ the specific suppliers of the domestically-designed chips, Huawei’s ‌Ascend chips are known to be integral⁤ to the ⁣country’s AI ambitions. Huawei is advancing in the AI space with its latest AI processor, the Ascend ‌910C, which is‍ currently being ⁣tested by large Chinese​ server companies and internet giants.

In its quest for technological independence, China Telecom is⁣ also partnering with other domestic chipmakers like Cambricon. These collaborations ‌are part of‌ a broader movement to establish a self-sufficient ecosystem‍ of AI solutions in China, insulating the country from ⁢the impact of US export⁤ controls.

As ‍China continues to ​develop its own ⁤AI chips and⁣ technologies, it is gradually lessening its ‍dependence on foreign hardware. Despite the challenges ⁢posed by US ‌sanctions and ‌the existence of ‌a black market for foreign chips, Chinese companies are opting for legitimate, albeit⁤ less powerful, alternatives to maintain access to official support and ⁢services.

Further Reading and Image Credits

For ​more detailed‌ insights⁤ into China⁢ Telecom’s AI advancements and the broader implications ‌for the global AI landscape, ⁢readers ‌are encouraged to view the original article. Image credit goes to ‍ Mark Kuiper.

Read the full article here.

Leave a Comment