Re: Concerns/questions around Software Heritage Archive

[Ludovic Courtès]

Hi,

Ian Eure <ian@retrospec.tv> skribis:

> They appear to be using the archive to build LLMs:
> https://www.softwareheritage.org/2024/02/28/responsible-ai-with-starcoder2/

To me, if the end result is that copyleft licenses are ignored, as is
the case with Microsoft’s CoPilot, then we have a problem.

That’s no excuse, but the problem goes beyond SWH: people upload copies
of repositories to GitHub without one’s consent (nothing to blame them
for, it’s free software), and then code ends up being used as training
data for CoPilot.

As you may have seen, this is being discussed on the Fediverse.  I’d
like to leave the SWH people time to reply to concerns that have been
raised.

> I was also distressed to see how poorly they treated a developer who
> wished to update their name:
> https://cohost.org/arborelia/post/4968198-the-software-heritag
> https://cohost.org/arborelia/post/5052044-the-software-heritag

That’s another concern, with append-only storage in general, starting
with Git.  We should look for solutions that work for both contributors
who change names and for users.  This has happened several times in Guix
and what people did was search/replace their name and adjust ‘.mailmap’.

Thanks,
Ludo’.