Keep Your Vibe Coding Out Of My Open Source Project

Key Takeaways

whether you like it or not often times I lean more towards the not myself AI text AI images video audio and of course code generation is here it's getting better and better every single day and if you think you're seeing less of it that's because in a lot of cases it's gotten so good that you don't actually realize that this is AI generated a lot of people still think there are these very obvious tells and for some things there certainly is but for a lot of it it's basically indistinguishable from Human work now whether that's a good thing or not that's totally up for discussion and I know there's a lot of people that value the artist over just a creation they want to see something made by an actual person and I completely understand that but there is a problem that has been getting bigger and bigger and it's a problem that you can't put your head in the sand about and just not think of we are seeing people grow up using AI tooling from a very young age people who are first learning to program first learning to draw first learning to do a lot of this stuff using AI tooling from the very beginning it is becoming more common to use this as a learning tool more common to use AI over using a search engine which to me seems absolutely insane I don't understand it but people are doing so so in the world of open-source software development the question becomes how much do you or don't you allow in your project a little over a week ago I came across this Lou no longer allows generative AI contributions this is the gnome image viewer typically just called image viewer but it actually does have a name I just merged in addition to lose contribution guidelines that bans the use of AI generated content in contributions for now I came up with the following text I am in favor of adopting a similar policy for all official gnome software use of generative AI this project does not allow contributions generated by large language models l M and chat Bots this ban includes tools like chat GPT Claude co-pilot deep seek and Devon AI I would say include but not limited to because of course there's going to be someone who tries to sort of circumvent the rules now if they're circumventing the rules you can ban them anyway but just to make sure you don't have to deal with that conversation includes but not limited to but it's a very minor nitpick we are taking these steps as precaution due to the potential negative influence of AI generated content on quality as well as likely copyright violations as I said at the start of the video I think quality is becoming a lot less of a concern but this is a problem you're never going to really deal with at least I don't know if anyone working on this problem copyright violations not following software licenses yeah that's a big issue and that by itself is enough reason to ban it this ban of AI generated content applies to all parts of the projects including but not limited to code documentation issues artworks and exception applies for purely translating text for issues and comments to English so if you are not a native English speaker and you maybe don't understand some of the wording being used in an issue or you're not really confident in your ability to write English and you want to use a translation to to go from whatever your native language is into English that's okay and I don't think there's anybody who would argue against that yes these tools are not perfect but this is a case of making things more accessible it's not a case of making issues which are just frankly gibberish you know finding things in the project which is simply not real this is making sure your text is somewhat understandable to a person in a different language and gnome like most F projects is English first so even though it is a worldwide project you kind of need to know English or have some way to deal with English if you want to get involved AI tools can be used to answer questions and find information however we encourage contributors to avoid them in favor of using existing documentation and our chats and forums since AI generated information is frequently misleading or false we can't cannot Supply support on anything referencing AI output this is understandable it's not something you wrote so of course you can't you know support that thing now when it comes to misleading information this is especially true in cases where there is a clear cut off point so if you want to know something about maybe the newest version of gome and that's not included in the data set well that's not included in the the data set also when there are cases of limited information about a thing being included in the data set it will fill in the gaps in many cases with things that seem reasonable enough but if you dig far enough well you know it's all a matter of how much data is there to actually build the output off of if it's something where there's like a little bit you're probably going to get nonsense if there's a lot hey maybe you might get something of use now regarding this point here about adopting it in other gnome software this is something that has been done following this being made now you might notice these are making sure you're not a bot The Gnome gitlab has added a thing where they do like bot detection now so it actually breaks every single link that is posted which is good but a bunch of projects in The Gnome Suite have adopted this exact text now this text is very heavily inspired by what Servo has over in their text it's basically the same idea here but they have the not limited to I don't know why you would base it off of that and then not include that very important part but I'm sure there's no I don't think there's a reason I think I just forgot to include that part and then include some reasons for why they do this maintain a burden correctness and security copyright issues and ethical issues these are two that are very much not going to be fixed because the companies involved are either don't care or don't think they are problems this is a matter of time and maintain a burden you know sort of depends on the specific contribution now gnome here is just an example Servo would work just as well or any other project that puts in a restriction like this the ultimate question then becomes how do you know that a cont contribution is AI there are a lot of people out there who think they are special who think they'll always be able to tell there's just something about that it's given me AI Vibes I know that's AI I don't have any evidence it's AI but there's just it it feels like AI now I'm not one of these people who think that at all and I'm not one of these people who think that AI cannot perform well or outperform especially the average or sub average developer because frankly not everybody is this amazing developer who creates the best code you'll see a lot of people out there are kind of just average that's why it's the average and when there's an average there's a lot of people below the average so once all the very obvious tells like overc commenting using fake libraries using old library things like that are resolved and the result you get is basically one contribution from a person and one contribution from an AI and they're basically indistinguishable it's following all the rules of the project the code works it does the thing the person says it does without any additional things when it comes to that point how do you know the contribution is AI and at that Point really all it comes down to is trust and Vibes do you believe the person you're dealing with is acting in good faith and this will work well enough for an established contributor or someone who has a history of contributing to things in the fuss world but I have some major concerns with how I've already seen that being handled the best example of this is over in artist communities if you spend any length of time on social media around people who draw you're bound to come across someone accused of using AI now it might be someone you know might be some random person but you're going to see accusations just fly around all over the place this is a very big deal in the art world because if you're accused of using Ai and you're using AI you're basically Satan now there are some people who from the start say they're using AI but those are people you know who've made their you know where they stand clear from the start but if you say that I made this and it turns out an AI made this well that's a pretty big deal you're going to get harassed you're going to get spammed with hate mail people are going to demand that you prove that you're not using AI now for some of these people for a lot of these people turns out yeah they were using AI especially early on where the toes were pretty obvious and even in cases where people were cleaning things up they would miss a lot of things and yeah okay you're using AI maybe you don't deserve all the hate you got right I think it's way too extreme how people handled but be honest if you're drawing it you draw it if you didn't draw it don't say you draw it that's weird that's cringe but then there are the people who turns out they were never using AI they've never used AI it's just their style is a very common style and a style that is a common model output but the result of the accusation is exactly the same they get harassed they get spammed with hate mail people demand to show that you're not using AI show me some proof and even when they show proof people still want more proof because people smelt blood in the water they found people that they could reasonably harass at least from their perspective so they are going to do it in the programming World opinions on AI aren't that strong but I'm still concerned with how the purity testing is going to go here because if someone produces code which kind kind of looks like AI code they're going to be treated as if they wrote AI code and they'll have to prove that they're not doing so now there is Tooling in place to make this easier in the Foss world we have things like GS off people are committing things over time then you can prove that but if you're not using Source control until you send it up to the project you know you're going to have a single commit there with all the changes being made and as these AI tools get better and better and there's more integration into development environments I just worry with how this is going to be enforced right like if someone has obviously proven to be using AI sure but if you're in a case where someone just throws it out there you know this can be very damaging to your position in the project now I do think intent is very important here a great example of this is with curl or any other big project that has a bug Bounty program so Daniel Stenberg the founder creator of KL will often publicize these bug reports where they're just nonsense they're the most generic thing you'll ever see just basically including the Wikipedia definition for buffer overflow and things like that or security vulnerability and if they do reference code it's either going to be code that doesn't exist in the code base code that used to exist in the code base or a mix of the two and these are people who are not using these tools to find a vulnerability these are people who see the output it says there's a vulnerability here they don't go and test it they don't go and verify it they just go and report it because there's a bug boun program because they think if this is accepted if it actually did find a real bug they're going to get paid for it this is someone who is acting maliciously this is someone who is typically using a boner account who is just opened it solely for the sake of this one issue these are people who rightfully should be removed from your project they are not trying to be a good contributor but that is a very different case from someone who maybe uses the cursor code editor who generates some code who modifies it a bit by hand who verifies that the thing that was generated actually does the thing that it says it does it actually fix a problem it actually introduces a feature and that person is using the AI tool as more of a assistant rather than as the entirety of the work this is very different from someone generating art and then pushing it as their own it's very different from someone generating a fake bug report as a way to try to get a bug bounty in the end you may still consider these bad and they still do have licensing issues they still have copyright issues and if you think it's a big issue they still have ethical issues where a lot of these models most of them all of them maybe not all of them most of the models are trained on all manner of code with absolutely no expectation of it's going to generate code that actually fits into a project it's not going to mix and match code bases that have completely conflicting licenses it's not going to use proprietary code that should not be made public that happens to be on a private account Microsoft they've done that a couple of times um sometimes their own code which is very funny when that happens that's also a big issue but a lot of people want to look at AI as this black and white issue where there's a right way and a wrong way to look at it but it's considerably more complicated than that personally I do think this is a good approach for the project to take I am concerned with the enforcement of this as things get better and better but from a legal liability standpoint you basically have to do this but what do you think do you think this is something that is a concern in the Foss world do you think War project should do this do you think Les should do it how do you approach this AI tooling do you use it yourself do you try to avoid it do you really think about the licenses your project use when you're dealing with this tooling I would love to know so if you like the video go like the video If you really like video and you want to become one of these amazing people over here check out the patreon SCB baray Linked In the description down below that's going to be it for me and hey [Music] something this