Caltech Researchers Claim Compression of High-Fidelity AI Models(wsj.com)

Caltech Researchers Claim Compression of High-Fidelity AI Models

https://www.wsj.com/cio-journal/caltech-researchers-claim-radical-compression-of-high-fidelity-ai-models-e66f31c9

3 comments

105 comments: https://news.ycombinator.com/item?id=47593422

“Show HN: 1-Bit Bonsai, the First Commercially Viable 1-Bit LLMs” (prismml.com) | 227 points | 10 hours ago

grimm80803 ay önce

I need a subscription to read that. Though, I don't get why 1 bit LLMs are smaller, if you go from 8 bit to 1 bit, you'll need 8 times more nodes right?

jonbaer3 ay önce

https://archive.ph/XIYG5