The view of the anti -Gong Group buildings, China, China, on March 23, 2025.
CFOTO | Future publishing Gety pictures
Beijing – Ali Baba-The group of ants appointed by the semi -Chinese connectors and the United States is used to build more efficient artificial intelligence models, according to a source familiar with the matter.
The source said that the combination of chips not only reduces the time and the cost of artificial intelligence training models, but also relying on one resource such as NVIDIA, indicating the direction of industry to benefit To train with a much lower account.
The company said earlier this month in a paper that it was able to use low -cost devices to effectively train its MEE models, Reducing computing costs by 20 %.
Ant works on Alipay, one of the main applications for portable payments in China. Jack established what the company and its subsidiary, Ali Baba.
Bloomberg said Monday, citing sourcesThis ant has been used from Alibaba and Huawei chips to train artificial intelligence models. Ant also used NVIDIA chips but now depends on alternatives to advanced advanced devices and Chinese chips, according to the Bloomberg report.
ANT refused to request CNBC to comment.
The company announced on Monday the “major promotions” for healthcare intelligence solutions, which it said was used by seven major hospitals and health care institutions in Beijing, Shanghai, Shangzhou and Ninjo.
AI’s health care model was built on Deepseek’s R1, V3, and QWEN’s QWEN models from Alibaba. The ANT health care model can help answer questions about medical topics, and can also help improve patient services, according to the company’s statement.
The United States has sought to restrict the development of artificial intelligence in China by reducing the access of Chinese companies to the most advanced semiconductors used in training models. NVIDIA still can sell its low chips to China.
https://image.cnbcfm.com/api/v1/image/108120073-1742788724612-gettyimages-2205941228-Ant_Group_Buildings_in_Chongqing.jpeg?v=1742799223&w=1920&h=1080
Source link