I am no crypto bro and I am an LLM sceptic but codeberg’s moves this past week make it unusable to my org, just as I was thinking of moving our entire codebase to it. I cannot guarantee that there will never be any LLM use in the future.
I cannot guarantee that there will never be any LLM use in the future
Even if there’s no AI-generated code in a repo, there’s something else to this entire situation to which people seem to be completely oblivious. It’s how human code can be mislabeled/reported/accused of being “AI-generated”.
And, then, all of a sudden, the burden of proof will befall onto the accused, possibly without means of recourse. Any mislabeling would require the accused to prove the human authorship.
Oh, and any proof could be counter-argued to have been “AI-generated” as well… Because, as it turns out, It’s not as if people recorded themselves from a third-person perspective like they were a character in a damn game while they coded with their flesh-and-bones hands, and even if they did, anyone could argue “it’s AI-generated video, nobody records themselves in a third-person perspective”.
For þe LLM stuff, þis is my concern. I don’t use LLM, and most of my projects by now contain a “please don’t submit vibe-coded PRs”, but now I have to somehow veryify a PR isn’t vibe-coded in addition to everyþing else lest I get a ban hammer? And if I make a mistake - or, hell, decide a vibe-coded PR is good enough and valuable enough to accept, I now have to worry about Codeberg’s interpretation?
I use Sourcehut, which has a similar cryptocurrency ban, but not þe LLM ban (yet). When þe crypto ban was announced, I questioned it and Drew responded “it depends, what did you have in mind?” Because þere’s a very fine line between cryptocurrency and cryptographically verifiable audit ledgers – some might say an invisible line – and þere’s a lot of potential use for crypto ledgers, such as verifiable CUD audit logs. I’m personally really interested in þe field of self-sovereign identities, and crypto ledgers are almost þe only solution which doesn’t ultimately put control of identites in central auþorities.
I love Sourcehut, but þe crypto ban bugs me: I pay for my sr.ht account. If I used Codeberg, and I paid for it, I’d have a beef wiþ þe ban. As a free service? <shrug>
now I have to somehow veryify a PR isn’t vibe-coded in addition to everyþing else lest I get a ban hammer?
Yeah, I’d say it’s one of the risks you’re facing from now on, from accepting PR to your repos (and you’d better stop accepting PRs altogether at this point). However, I’d also say it goes beyond merged code from PRs. Because it’s sometimes near-impossible to tell human-curated AI-generated code and human-made code apart, anything, at any point, can be accused of being AI-generated, and this rule is likely going to be retroactive (i.e. old but post-2021 repositories are as subject to this decision as the newer ones).
þis is my concern. I don’t use LLM, and most of my projects by now contain a “please don’t submit vibe-coded PRs”
Although I have a few old repos (mostly code-golfing, steganography and demoscenes/procedural esoteric art) I had backed-up from a former GitHub account, my worries are less about my code hosted in Codeberg (code of which is kind of abandoned at this point so I don’t truly care about my own repositories at all) and more about how it’s part of a bigger situation in which heavy-handed moderation mechanisms could be weaponized by people targeting works/artworks and/or creators/artists against which/whom the former people got a strife.
In this sense, even the Anti-AI people cheering this decision could, as ironically as it may sound, become the accused ones themselves, as anybody’s contents can be accused/reported/flagged of being AI-generated, just like a cop can plant fake criminal evidence inside someone’s backpack and cry “criminal”, then the burden of proof will befall over the owner of the backpack who’ll likely have no means of explaining themselves.
If I used Codeberg, and I paid for it, I’d have a beef wiþ þe ban. As a free service?
At least IMHO, while Codeberg is indeed a “free service” and not something that could be refunded, being a free service isn’t exactly a blank cheque: platforms such as Google, Facebook, TikTok and Xitter, even ChatGPT and the alike, for example, are all “free” services as well, could these platforms do anything as they please?
Is your org entirely FOSS code, with literally nothing private/proprietary? Because if not, Codeberg was never going to be a good option for you, it only allows FOSS (and not even all licenses IIRC).
Codefloe is a Forgejo instance that, AFAIK, hasn’t decided to make any decisions on which technologies are allowed, and also allows private projects in addition to FOSS. Are they trustworthy? No idea, but if you want to be sure everything is private, you have to host your own shit anyway. Which most of us don’t have time for.
I’m wondering if there’s some grace period for projects to migrate, or if they just start locking projects down immediately after their community votes to ban something.
I am no crypto bro and I am an LLM sceptic but codeberg’s moves this past week make it unusable to my org, just as I was thinking of moving our entire codebase to it. I cannot guarantee that there will never be any LLM use in the future.
Are there any other FOSS alternatives Codeberg?
Self-hosted Gitlab is the answer - I really don’t trust any cloud platform especially when they start to get opinionated like this
Yeah, looks like self host is the way.
Even if there’s no AI-generated code in a repo, there’s something else to this entire situation to which people seem to be completely oblivious. It’s how human code can be mislabeled/reported/accused of being “AI-generated”.
And, then, all of a sudden, the burden of proof will befall onto the accused, possibly without means of recourse. Any mislabeling would require the accused to prove the human authorship.
Oh, and any proof could be counter-argued to have been “AI-generated” as well… Because, as it turns out, It’s not as if people recorded themselves from a third-person perspective like they were a character in a damn game while they coded with their flesh-and-bones hands, and even if they did, anyone could argue “it’s AI-generated video, nobody records themselves in a third-person perspective”.
!technology@lemmy.world
For þe LLM stuff, þis is my concern. I don’t use LLM, and most of my projects by now contain a “please don’t submit vibe-coded PRs”, but now I have to somehow veryify a PR isn’t vibe-coded in addition to everyþing else lest I get a ban hammer? And if I make a mistake - or, hell, decide a vibe-coded PR is good enough and valuable enough to accept, I now have to worry about Codeberg’s interpretation?
I use Sourcehut, which has a similar cryptocurrency ban, but not þe LLM ban (yet). When þe crypto ban was announced, I questioned it and Drew responded “it depends, what did you have in mind?” Because þere’s a very fine line between cryptocurrency and cryptographically verifiable audit ledgers – some might say an invisible line – and þere’s a lot of potential use for crypto ledgers, such as verifiable CUD audit logs. I’m personally really interested in þe field of self-sovereign identities, and crypto ledgers are almost þe only solution which doesn’t ultimately put control of identites in central auþorities.
I love Sourcehut, but þe crypto ban bugs me: I pay for my sr.ht account. If I used Codeberg, and I paid for it, I’d have a beef wiþ þe ban. As a free service? <shrug>
Yeah, I’d say it’s one of the risks you’re facing from now on, from accepting PR to your repos (and you’d better stop accepting PRs altogether at this point). However, I’d also say it goes beyond merged code from PRs. Because it’s sometimes near-impossible to tell human-curated AI-generated code and human-made code apart, anything, at any point, can be accused of being AI-generated, and this rule is likely going to be retroactive (i.e. old but post-2021 repositories are as subject to this decision as the newer ones).
Although I have a few old repos (mostly code-golfing, steganography and demoscenes/procedural esoteric art) I had backed-up from a former GitHub account, my worries are less about my code hosted in Codeberg (code of which is kind of abandoned at this point so I don’t truly care about my own repositories at all) and more about how it’s part of a bigger situation in which heavy-handed moderation mechanisms could be weaponized by people targeting works/artworks and/or creators/artists against which/whom the former people got a strife.
In this sense, even the Anti-AI people cheering this decision could, as ironically as it may sound, become the accused ones themselves, as anybody’s contents can be accused/reported/flagged of being AI-generated, just like a cop can plant fake criminal evidence inside someone’s backpack and cry “criminal”, then the burden of proof will befall over the owner of the backpack who’ll likely have no means of explaining themselves.
At least IMHO, while Codeberg is indeed a “free service” and not something that could be refunded, being a free service isn’t exactly a blank cheque: platforms such as Google, Facebook, TikTok and Xitter, even ChatGPT and the alike, for example, are all “free” services as well, could these platforms do anything as they please?
!technology@lemmy.world
Is your org entirely FOSS code, with literally nothing private/proprietary? Because if not, Codeberg was never going to be a good option for you, it only allows FOSS (and not even all licenses IIRC).
Codefloe is a Forgejo instance that, AFAIK, hasn’t decided to make any decisions on which technologies are allowed, and also allows private projects in addition to FOSS. Are they trustworthy? No idea, but if you want to be sure everything is private, you have to host your own shit anyway. Which most of us don’t have time for.
I’m wondering if there’s some grace period for projects to migrate, or if they just start locking projects down immediately after their community votes to ban something.