Delay in the launch of a novel chatbot model by DeepSeek attributed to hardware issues with Huawei processors.
In August 2020, Chinese authorities banned the use of Nvidia's H20 AI chips due to security concerns and potential backdoor fears [2][3]. This decision had a significant impact on the development of AI projects in China, as shown by the case of DeepSeek's R2 chatbot.
Major Chinese tech firms, including ByteDance, Alibaba, and Tencent, were ordered to cease purchasing Nvidia’s H20 AI chips, citing the chips as "not safe" for Chinese users [2]. This move reflects China’s broader push to reduce dependence on US technology amid geopolitical and security disputes.
Although the exact consequences for DeepSeek's R2 chatbot model are not explicitly detailed, the restriction likely affected its development by limiting access to key GPU technology needed for advanced AI training and deployment. AI models like chatbots rely heavily on powerful GPU technology, and a halt in using Nvidia chips in China during this period could have hampered the access to advanced AI processing hardware, likely causing delays or forcing reliance on alternative chips or architectures.
Despite these challenges, the delay in releasing the R2 chatbot model in May was reportedly due to difficulties in using Huawei chips for model training [4]. The Chinese authorities had advised the R1 chatbot model to use Huawei's Ascend processors instead of Nvidia chips. However, Chinese authorities sent a team of engineers to DeepSeek's office to address the issues with Ascend processors, but the Chinese company was still unable to use them for model training.
Currently, Huawei chips are only used for generating chatbot responses, while Nvidia chips are used for model training in the development of R2. The use of Nvidia chips for model training was necessary due to the technical difficulties encountered with Huawei's Ascend processors.
The US government's deal with Nvidia and AMD, under which the companies would pay 15% of their AI chip sales revenue to the US government in exchange for licenses to supply their products to China, may have influenced the readiness of US authorities to allow chip supplies to China [1]. However, the exact impact of this deal on the availability of Nvidia chips in China during the development of R2 is not clear.
In conclusion, the ban on Nvidia H20 chip use in China had a significant impact on the development of AI projects, as demonstrated by the case of DeepSeek's R2 chatbot. The restriction likely caused delays or forced reliance on alternative chips or architectures due to the limited access to advanced AI processing hardware. As more specific details on the development timeline or alternative hardware strategies become available, they would clarify the exact consequences further.
References: 1. The New York Times and FT reported that Nvidia and AMD had reached a deal with the administration of US President Donald Trump. 2. Chinese authorities urged companies to stop using American Nvidia chips, specifically the H20 AI chips, in August 2020. 3. This demand came amid escalating trade tensions and scrutiny over hardware security between China and the US. 4. The delay in releasing the R2 chatbot model in May was reportedly due to difficulties in using Huawei chips for model training.
- Due to China's ban on Nvidia's H20 AI chips and the subsequent need for alternative technology, major tech companies like ByteDance, Alibaba, and Tencent may have faced challenges in their AI projects, particularly in areas requiring powerful GPU technology for advanced AI training and deployment.
- The impact of China's decision to restrict the use of Nvidia's H20 AI chips extends to the development of advanced technology like artificial-intelligence models, such as chatbots, which rely heavily on such technology for their Deployment and training processes.
