HACKER Q&A
📣 akmittal

Should open source allow AI training?


Open source has enabled developers to learn new patterns, build companies for almost free and has made a huge positive impact on developers.

Recently we have seen generative AI such as ChatGPT trained on open source code which are so good at programming. It has made us concerned about loosing/impacting our jobs. Should open source allow AI companies which plan to make billions using their work, to train on open source code?

  👤 shagie Accepted Answer ✓
https://opensource.org/osd/

    6. No Discrimination Against Fields of Endeavor

    The license must not restrict anyone from making use of the program in a specific field of endeavor. For example, it may not restrict the program from being used in a business, or from being used for genetic research.

    ...

    10. License Must Be Technology-Neutral
    
    No provision of the license may be predicated on any individual technology or style of interface.
I'm not sure how one can claim to be open source and not allow for it to be used for AI training.

There are certainly license / fair use / derived work issues to be resolved... but the "allow AI training" probably should be a "yes" for open source software.

👤 maximejkb
I understand the concern and think it can even be broadened. Generative AI also threatens the pay and livelihood of graphic designers, artists, journalists, copywriters -- people who tend to be far more poorly paid than tech workers already. This application of generative AI to code-writing is really just automation coming home to roost. The question is, why should software engineers represent some special protected class from the dangers of automation?

I guess my point is there has to be some consistent standard here. Software engineers have no greater claim to their public-facing work than writers and artists, and are paid far more. If there's an argument to exclude open-source code from training sets there should also be an argument to exclude essays and journalism, digital drawings, and music from such training sets.