Huawei's DeepSeek-R1-Safe: Revolutionizing AI with Safety in Mind
Huawei has developed a new AI model, DeepSeek-R1-Safe, as a safer alternative to existing models. The model boasts a near-perfect success rate in avoiding politically sensitive discussions. This development highlights China's effort to ensure AI reflects socialist values, with stringent controls on speech.
In a groundbreaking collaboration, Chinese technology titan Huawei has unveiled DeepSeek-R1-Safe, a safety-centric rendition of its AI model, in partnership with Zhejiang University. The model reportedly achieves near-total success in circumventing discussions deemed politically sensitive by Chinese regulators.
The initiative is part of a broader regulatory mandate for domestic AI models and their derivatives to embody China's socialist principles. Huawei revealed that 1,000 Ascend AI chips were deployed to fine-tune this large-language model, derived from DeepSeek's open-source R1 variant.
This announcement comes as Huawei hosts its annual conference in Shanghai, ending a prolonged period of secrecy around its chipmaking prowess. The tech behemoth is poised to launch new chip and computing power roadmaps, marking a significant milestone in China's AI strategy.
Google News