ketchup32613

3 ms·

1. 1.▲
   If LLMs Have Human-Like Attributes, Then So Does Age of Empires II(arxiv.org)

   118 points by ketchup32613 4mo ago|121 comments
2. 2.▲
   by ketchup32613 4mo ago

   Do you want to see scaling curves wrt data and param size? I agree that 1.2B and 10B tokens is not representative, but what scale of parameters and dataset sizes would be convincing?
3. 3.▲
   Majorana 2 [video](youtube.com)

   3 points by ketchup32613 4mo ago|0 comments
4. 4.▲
   Apple M1 chip deep dive [video](youtube.com)

   4 points by ketchup32613 4mo ago|0 comments
5. 5.▲
   by ketchup32613 4mo ago

   You're not wrong about that. Speculative decoding does not affect the quality of tokens generated, as each token has to be verified by the parent model before it is output. Each of the tokens generated by the draft model has to be veri
6. 6.▲
   Nvidia introduces back end for CUDA kernels in Rust(github.com)

   3 points by ketchup32613 5mo ago|0 comments
7. 7.▲
   Hardest Hard Drive Commercial [video](youtube.com)

   2 points by ketchup32613 7mo ago|1 comments
8. 8.▲
   NemoClaw, "Enterprise" OpenClaw Implementation(nvidia.com)

   11 points by ketchup32613 7mo ago|0 comments