Datasets:
Dutch version
Just putting this here so it is easier to find for others: I translated the wonderful Dolly dataset with gpt-3.5 to Dutch to, hopefully, accelerate research for Dutch in the LLM world. It can be found here: https://huggingface.co/datasets/BramVanroy/dolly-15k-dutch/
Are you able to retain the CC BY SA 3.0 license if the output is from OpenAI? Seems like the terms of use aren't quite the same
Are you able to retain the CC BY SA 3.0 license if the output is from OpenAI? Seems like the terms of use aren't quite the same
Good point - last I read about this was that the license can be the same but that OpenAI's ToS specifically prohibit the use of their generated to build/launch competitive products. So not as strict as a fully NC license but more restricted to one company. I will investigate whether a non-commercial license is available that is compatible with the original.
EDIT: I have since written/asked about this here and on Twitter. In all my synthetic datasets I have added a reference to OpenAI's policies and Terms of Use. As far as I am aware they are not as restrictive as a full NC but are very specific to one use-case (as described in the SE Legal post). This has also been explained and added in the READMEs of my datasets. I hope that that suffices for now, and if not I hope that OpenAI gets back to me on this and if so I will updated this post/the licenses where possible, or remove the datasets from the public if needed.