Can proposing a fix to code with llm generated context prevent an issue from being fixed
So I play an open source game called Cataclysm: dark days ahead , and I had my save get bricked over some jank. I used llms to unbrick my save (I compiled the code put my save in and I was able to continue). So I did fork request and disclosed that I used llms to generate the code just to be honest and respect the people that care about llms.
So since I used an llm articulated solution for the problem and the community doesn’t allow llm content could that pull request be used an an excuse as a derivative source for the problem that bricked my save?
The rest is just a rant. I did go to the discord and Post that I made a pull request and I wanted to make sure I following procedure. I disclosed that I used LLM in my pull request and the user proceeded to get hostile, like actually triggered. I have to defend myself for my disclosure and they try to say I’m violating a discord rule with my disclosure, even though the rule is that I can’t post ai generated content in the discord. Then they demand the subject get dropped after they escalate.
I probably needed to get disillusioned over a game I like so much. I’ll put the game down and move onto other things.
Comments
Are you a developer? If not, imagine a layman coming to your pet project and saying they fixed something with an LLM. Would you:
You forgot the fourth bullet,
Optional :)
If you open an issue and link the PR with the code modifications, it will be useful for other people even if it does not get merged. After you do that you can just ignore everything.
That is not useful, when every layperson submits a fix for their issue and code review needs to be performed on all of them, suddenly it’s hundreds of solutions all with different code styles and likelyhoods of introducing their own bugs. This shit is driving open source out of existence. Don’t submit patches if they don’t want LLM code god dammit, you people are biting the hand that feeds
I checked the contributing.MD file and they do not allow LLM code. I still would open an issue and a link to your fork so people with the same issue can fix it. If the fix can be done by editing the save file directly I’d create a repo with the script instead.
If you ask me, a fix is a fix if it works. Correct code doesn’t suddenly become incorrect just because it was spat out by an LLM. That said, communities are free to establish boundaries regarding LLM use, and you did the respectable thing and disclosed the usage allowing them to make an informed decision.
You can also make the patch available on your own site with the same disclosure and let others benefit if they choose. If nothing else, it will put pressure on the devs to develop a conventional fix which complies with the rules and makes everyone happy.
Normally, forcing a dev to rush a fix or risk splintering their audience is a bit of a dick move, but based in the comments here, it sounds like dicketry is already afoot, and there isn’t any moral high ground to be had.
That game is one of my favorites of all time but the one time I interacted with the dev team 7 years ago I came away annoyed. Don’t meet your heroes.
I had that same thought after the discord discord (I’m making a pun). They do come off as hyper online, Reddit mod wannabees.
If its an open source game you can likely fork the entire thing with your fix and keep playing it.
True and only complaint is that it takes like an hour to compile the game, if I wanted my fix in newer versions
As much as I resent LLMs for seemingly taking away my career, a fix is a fix, and I find rejecting your fix just because you used a LLM at one point to be futile.
Ehhh, technical debt is a thing and just because something fixes an issue doesn’t mean it’s code you want to have to maintain in perpetuity.
I should say the GitHub does say no llm based pull request.
I do suspect that contributer probably inserted llm code simply because they didn’t want to type out repetitive code.
I wanted to be proactive about finding a solution to my bricked save and sharing the solution.
I do have the mindset that capitalism is the problem not llms and llms are a tool
Depending on the repo, they might be more annoyed about completely automated pull requests. As long as you write the pull request manually, are clear that you are a human, but you used AI, you may have some luck.
Yeah, if you can explain the fix there’s no reason to reject it. If there’s 7k lines changed it’s going to get rejected. If it’s a clean 50 that fixes an issue it should be accepted. Doing PRs properly takes time and LLMs try to change more than needed
Sign in to comment. Sign in