After what would only be described as an action-packed week for DeepSeek, the Chinese AI company isn’t done making waves in the tech industry.
DeepSeek shakes up the AI industry
DeepSeek caught the tech world by surprise with the release of R1, their large language model, to rival OpenAI’s ChatGPT. The R1, which was praised by western tech companies for its performance, not only matches but even surpasses ChatGPT in some benchmarks despite having a budget of only $5.6 million.
U.S. sanctions foiled attempt to stunt China’s AI growth
That’s also with the sanctions that the U.S. government has placed on China that restricted semiconductor development and access to advanced chips that western AI companies use to develop and train their AI models. DeepSeek’s arrival seems to question the sanction’s ability to curb AI development in the east.
DeepSeek turns to HUAWEI for AI processing
The R1 was initially trained on NVIDIA’s H800 chips, but recent reports have revealed that the company is using HUAWEI’s Ascend 910C chips to run inference (AI’s ability to process newly received information). Despite falling a bit short of the H800’s performance, DeepSeek is still considering running their next model on the Chinese-made chips.
A strategic shift towards western independence
It seems that the move to use HUAWEI chips isn’t just a practical one; it can also be considered as a power play to break away from western dependence on tech development.