4 ms·

> DeepSeek built their base models off Llama releases and OpenAI outputs Those models are also trained on data that was ignoring licenses / copyrighted content

by timeon 2y ago

> DeepSeek built their base models off Llama releases and OpenAI outputs Those models are also trained on data that was ignoring licenses / copyrighted content.

ithkuil 2y ago
That's a problem for sure. But why would that argument play in favour china which would be even less constrained by licenses / copyright ?
- timeon 2y ago
  I didn't mean it as favour for China. Just that this industry is unfortunately like that.