3 ms·
I think the point is - OpenAI scraped public data - d1 - Trained their model to produce output - d2 - DeepSeek used d2 to reinforce their model OpenAI is mad
by the_arun 2y ago
I think the point is
- OpenAI scraped public data - d1
- Trained their model to produce output - d2
- DeepSeek used d2 to reinforce their model
OpenAI is mad about d2 (not d1). I'm not sure using public data is "stealing". In summary, these are two different things & need to be separate.
- redleader55 2y agoYou say "public", but what I think you mean is "publicly available". Even publicly available data has copyrights, and unless that copyright is "public domain", you need to follow some rules. Even licenses like Creative Commons, which would be the most permissive, come with caveats which OpenAI doesn't follow [0]. It is unclear if someone breaking someone else's copyright to use A can claim copyright on a work B, derived from A. My point is that OpenAI played loose with the copyright rules to build its various models, so the legality of their claims against DeepSeek might not be so strong. [0] https://creativecommons.org/share-your-work/cclicenses/ https://creativecommons.org/share-your-work/cclicenses/
- the_arun 2y agoI am not saying OpenAI did good by using publicly available data. I meant these are separate activities. None is good. But DeepSeek is slightly better by making theirs opensource.
- xbar 2y agoOpenAI (sc)raped all the data it could. I do not accept your assertion that d1 was "public." It was accessible, for certain. OpenAI asserts 1. d2 was used by DeepSeek 2. All d2 belongs to OpenAI exclusively Both are debatable for large number of reasons.