IT giant HCLTech CEO C Vijayakumar said that India needs to build its own large language models (LLMs) to limit reliance on open-source AI models and have a long-term competitive advantage ...

以FutureLabs首席专家胡延平为代表的一些人认为，DeepSeek的亮点，按实际价值其实依次是强化学习、混合专家模型MoE、知识与模型蒸馏、多头注意力、多Token预测、混合精度训练和PTX等。尽管其中没有一项是DeepSeek自己完全原创、独 ...