GrapheneOS stance on LLM generated content

DeletedUser997

Does GrapheneOS have a stance on LLM generated content (code, assets etc.)? If so I suggest implementing an official policy (ex. secureblue, Fedora) to govern its use in the project.

precise-deliciousness

The pinned post by the official GrapheneOS project account, https://discuss.grapheneos.org/d/11951-ai-generated-text-is-forbidden-with-the-exception-of-automated-translation would imply their stance on generated content in general and rightfully so! I can't imagine it being any different for pull requests.

GrapheneOS

precise-deliciousness The policy posted there applies to our chat rooms, forums, issue tracker, replies to our social media accounts and explanations for pull requests.

People who aren't developers shouldn't submit AI generated issue reports or pull requests. People who are developers and fully confirm the output from the AI is correct with a full understanding of it can submit an issue report or pull request based on it. They're taking full responsibility for it and submitting it is an assertion they understand it and confirmed it's a real problem and/or fix for it. It's only acceptable for small things where the person submitting it can thoroughly understand and review it. We aren't going to accept large external changes in general so there's no point generating those with AI. It's a waste of our time especially since AI generated code won't come close to meeting our standards without major modifications.

MineralWater

Isn't AI generated code much extra work? It has to be reviewed and it is a security risk.

I mean, the owners of the AI could decide from one day to the other that they implement backdoors or fragments of these to the generated code. Or the AI runs mad and implements something harmful.

AI hasn't shown yet that it can be used totally risk free and in the end it is just a tool to make people dependent on AI companies.

Just like for example Google, who killed their search with AI and wants to present only AI answers, no more web pages. So in the end what is presented as a result could be fully fictional, as you never know if it was hallucinated or drawn from web pages or a mix of both.

People always get fooled with the belief that the AI is their buddy and let's them work faster. No, it lets you unlearn things, because you don't practice anymore. Like Google Maps made most people totally incompetent in reading real maps and contact apps made them unable to remember complete phone numbers, even their own.

GrapheneOS

Removed the off topic discussion not belonging in the Development section.