Re:An analysis of DeepSeek's R1-Zero and R1
by danbala on Thursday January 30, 2025 @04:01PM
(#42882129)
shouldn't the whole idea be: get away from needing data at all? if a model can really reason, it should be able to figure things out on its own.
[ Parent
| on HN ]
Re:An analysis of DeepSeek's R1-Zero and R1
by FuriouslyAdrift on Thursday January 30, 2025 @04:20PM
(#42882307)
An LLM is just a really good parser connected to a lossy compressed corpus of data.
They need to be open ended and self training to be truly useful.
Reasoning is way far away...