Conversation
Notices
-
Embed this notice
I don't think Linux-libre as a project would set out to clean up LLM-generated code from Linux.
not only because that's not necessary to attain the goal of a freedom-respecting kernel, but because identifying and removing such code would probably make the project unsustainable.
finding embedded blobs and blob requests turned out to be reasonably easy to automate, despite the large number of false positives that require manual intervention.
I don't see that recognizing LLM-generated code is automatable. heck, I'm not even sure it can be done at all in general, whether algorithmically or intuitively.
it would be interesting if someone with a lot more time in one's hands were to set out to identify this kind of code and strip it out.
CC: @phf@dmv.community @devuan@toot.community @jaromil@mastodon.social @freedo@mastodon.social @civodul@toot.aquilenet.fr @guix@hachyderm.io
-
Embed this notice
@lxo @jbc @devuan @jaromil @freedo @civodul If you can find out and keep a record as to who develops with proprietary software and SaaSS to make slop, it's somewhat possible to detect automatically which code is proprietary.
Yes, software is rendered proprietary by random inclusion of a nontrivial amount of code under incompatible or proprietary licenses or even if the license is compatible, the non-inclusion of the necessary copyright information infringes all real free licenses (the LLM usually doesn't include that information and often puts the wrong license in if prompted for the license).
As proprietary software doesn't respect the users freedom, a freedom-respecting kernel would need to remove such proprietary software.
Linux may need to be dropped and Hurd finished at this rate.