DeepSeek V3.2 introduces Native Sparse Attention architecture designed to make long-context models more efficient without sacrificing performance. Instead of applying sparsity only during inference (after training is complete), NSA is designed to be sparse from the very beginning and is trainable from end-to-end. ... read full post
cross-posted from: https://lemmy.ml/post/36807834
China has issued a guideline to accelerate the integration of artificial intelligence into the transport sector, setting goals for widespread adoption by 2027 and deeper integration by 2030. ... read full post
China proposed the AI+ International Cooperation Initiative at a high-level meeting on the Global Development Initiative on Tuesday. Chinese experts said that China regards AI as a universal public good for the global community. ... read full post
Such a mechanism will ultimately make it possible to almost completely resolve the problem of radioactive waste accumulation, the Russian leader said ... read full post