NHK

Data-Decoding Method Using AI in Multi-level Amplitude Modulation Holographic Memory

As a next-generation optical archive storage device for 8K video, we have been researching and developing holographic memory. Recent studies have focused on multi-level code instead of binary code to ...

New KV cache compaction technique cuts LLM memory 50x without accuracy loss

MIT researchers developed Attention Matching, a KV cache compaction technique that compresses LLM memory by 50x in seconds — ...
MSN による配信

New AI method lets models think harder while avoiding costly bandwidth

DeepSeek’s Engram separates static memory from computation, increasing efficiency in large AI models The method reduces high-speed memory needs by enabling DeepSeek models to use lookups Engram ...

MIT Finds Way To Shrink AI Memory 50x Without Losing Accuracy

This breakthrough could make AI far more practical for large-scale use as the method promises to cut cloud computing costs and process huge datasets faster.