GCC Adopts Policy Rejecting Significant AI-Generated Code (linuxiac.com) 121
GCC has adopted a policy rejecting substantial code contributions generated by or derived from LLMs. "This covers not just code copied directly from tools like ChatGPT, Gemini, or GitHub Copilot, but also any versions of the code later edited or rewritten by a human, provided that the final contribution is still based on material generated by the system," reports Linuxiac. From the report: The important point, then, is not whether a developer has used an AI tool at some stage in their work; contributors can use LLMs to discuss ideas, understand existing code, learn about a field they are unfamiliar with, or carry out general research. The limitation lies in the inclusion of copyright-significant material generated by such tools in the code submitted to GCC.
The policy also provides for a few limited exceptions; GCC maintainers are allowed to accept changes that are legally insignificant or trivial and are generated by an LLM, on the condition that they meet the project's normal contribution requirements and the use of an LLM is clearly disclosed.
Furthermore, copyright-significant AI-generated test cases could still be accepted. Since test cases usually involve small programs which are intended to reproduce compiler bugs or to verify certain behavior, the policy deals with them separately from code that is incorporated into GCC itself.
It does not follow that merely looking at or altering the generated code makes it acceptable. By the rules that have been adopted, a contributor cannot take substantial implementation produced by an LLM, clean it up manually, and then treat the resulting patch as if it had been originally written by them. Once a contribution has been derived from generated content, it is still subject to the policy.
The policy also provides for a few limited exceptions; GCC maintainers are allowed to accept changes that are legally insignificant or trivial and are generated by an LLM, on the condition that they meet the project's normal contribution requirements and the use of an LLM is clearly disclosed.
Furthermore, copyright-significant AI-generated test cases could still be accepted. Since test cases usually involve small programs which are intended to reproduce compiler bugs or to verify certain behavior, the policy deals with them separately from code that is incorporated into GCC itself.
It does not follow that merely looking at or altering the generated code makes it acceptable. By the rules that have been adopted, a contributor cannot take substantial implementation produced by an LLM, clean it up manually, and then treat the resulting patch as if it had been originally written by them. Once a contribution has been derived from generated content, it is still subject to the policy.
Copyright (Score:5, Informative)
"The limitation lies in the inclusion of copyright-significant material generated by such tools in the code submitted to GCC"
This seems like the important bit. Isn't there significant concern over whether LLM generated code can be copyrighted?
Re:Copyright (Score:5, Insightful)
It has already been established that LLM code either has no copyright whatsoever or (worse) it retains the original copyright and ownership if it is too similar to things in its training data. This is a really bad legal minefield.
Re:Copyright (Score:5, Interesting)
Winner! Winner! Chicken Dinner!
Yep, even when AI generated code is human reviewed to determine that the code does what it looks like it should, and has no subtle issues (or only very very subtle ones!), you still have the huge legal issues (the exact ones you identified).
Re: (Score:3)
It's going to be very interesting. There are already some copyright owners that provide special models of their own which allow them to attribute percentage ownership from the training data to the generated output. They can do more if they have the full knowledge of the training corpus and details of the model weights, but they don't need that to make "educated guesses" which they can use to sue and then force access to the model that was used to generate the output.
Re: (Score:3, Interesting)
* Very few are going to win the "AI stole my work" lawsuits. They are shuffling deck chairs on the Titanic
* Even if some do it won't help others who will stay away from those fields unless truly hardcore. You going to tell your kids to be web designers or graphic artists now? Musicians? Programmers? Goooooooooooood luck with that. Tell them to go into buggy whip manufacturing, too, while you are handing out advice.
* After the 6th gen of LLM bots they got better than most coders. Mys
Re: Copyright (Score:2)
Getting stuff done quickly is always going to win over getting stuff done slower due to arbitrary constraints.
I've demonstrated the principle on a few open source projects where issues remained unaddressed for months, or where I was told "it can't be done". All done in a matter of days or hours.
I've hardly written a line of code for a year now. I don't expect AI to do things flawlessly, but I have a 5 digit slashdot ID and therefore the experience to know where a clever but naive unseasoned developer is lik
Appeal to authority fallacy (Score:3)
have a 5 digit slashdot ID and therefore the experience to know
That’s an appeal to authority fallacy. A 5 digit ID is not evidence of competence. There were plenty of low skill devs who participating in the earliest days of slashdot.
I've hardly written a line of code for a year now. I don't expect AI to do things flawlessly,
But what is the quality of that code? How easy is it to understand for the next human dev brought onto the project?
I’ve watched a coworker do as you describe. Use the AI not only for developing new code but to debug and update that code as bugs manifest. It was a bit of a mess. The AI routinely broke formerly working code. If
Re: Appeal to authority fallacy (Score:2)
Re: (Score:2)
Why would anyone care whether the next dev can work with it?
That’s the difference between professionals and amateurs. When a program will be of a certain complexity or a certain importance, you want the devs to be professional.
Re: (Score:2)
A 5 digit ID is not evidence of competence.
Definitely not. You need 4 digits for that. :)
But what is the quality of that code? How easy is it to understand for the next human dev brought onto the project?
I’ve watched a coworker do as you describe. Use the AI not only for developing new code but to debug and update that code as bugs manifest. It was a bit of a mess. The AI routinely broke formerly working code. If lost the context on why it had written the code that way in the first place, and when fixing a bug in that area it failed to consider a special case it no longer was away off. Introducing a new bug. I watched fixing one bug generate many iterations of one step forward one step back. The code suffered greatly as a result.
When I assisted him on a couple of occasions I was constantly spotting things both he and the AI missed. AI needs rigorous oversight for non-boilerplate code. Its a useful tool, but it can easily be misused. Often through overuse.
But that's the thing: All of that applies equally to human generated code, too. I've seen plenty of humans do all of those things even without AI assistance.
Where AI today consistently beats a human is on speed. No human can generate code as fast as an AI can. But speed of generation isn't everything. Right now (competent) humans have the edge when it comes to seeing the big picture, anticipating the code evolution over time as requirements change, and stu
AI - more specs and more tests before coding (Score:2)
But what is the quality of that code?
Not to the same degree. The junior dev is more likely to remember that special case the code must consider. The junior dev learns while designing and writing the code. The junior dev learns from their mistakes.The AI, not really. What it codes is still based on its training data.
How easy is it to understand for the next human dev brought onto the project?
Easier than the AI if they have some experience. Are past the junior stuff mentioned above.
I’ve watched a coworker do as you describe. Use the AI not only for developing new code but to debug and update that code as bugs manifest. It was a bit of a mess. The AI routinely broke formerly working code. If lost the context on why it had written the code that way in the first place, and when fixing a bug in that area it failed to consider a special case it no longer was away off. Introducing a new bug. I watched fixing one bug generate many iterations of one step forward one step back. The code suffered greatly as a result.
When I assisted him on a couple of occasions I was constantly spotting things both he and the AI missed. AI needs rigorous oversight for non-boilerplate code. Its a useful tool, but it can easily be misused. Often through overuse.
But that's the thing: All of that applies equally to human generated code, too.
To a lesser degree, and its self correcting. The junior dev l
Re: (Score:2)
Re: (Score:2)
Getting stuff done quickly is always going to win over getting stuff done slower due to arbitrary constraints.
GCC is largely complete. Moving quickly is not an important requirement, but maintaining stability IS an important requirement.
Re: (Score:2)
Re: Copyright (Score:2)
"I'd advise you you to shut the fuck up and work on a project that's now possible where before you needed a whole team."
Why, so they can replace me with someone cheaper next week? Or do you mean I should create some project in hope of a buyout when the organization that could buy it from me can just prompt their own version without me?
Re: (Score:2)
Re: (Score:2)
I'd advise you you to shut the fuck up and work on a project that's now possible where before you needed a whole team.
What project is that?
Sovereign system. (Score:2)
https://www.scry.llc/2026/04/0... [scry.llc]
Re: (Score:2)
Re: Copyright (Score:2)
GCC committing suicide.
Might as well move to Spain and wait for the Horde to overrun them.
Re: (Score:2)
Re:Copyright (Score:5, Insightful)
Indeed. Of course the usual mindless LLM cultists will claim this is all imaginary or not a problem or similar nonsense. Fortunately the GCC people are smarter. I expect many FOSS projects will go the same way as GCC in this. The risks are just far too large and the gains from using LLM code are, at best, questionable.
Re: (Score:2)
The problem is "How do you tell?". Certainly many things that claim to be evidence really aren't.
You can check that the code does what it claims to do. You can't really check that the claimed author really wrote it.
Re: (Score:2)
Most people are honorable and will not do it. The reminder (people of low quality) can at least be immediately and publicly kicked off the project when they lie. And that threat alone should scare them enough.
Re: (Score:3)
Re: (Score:2)
Yes. And the thin is even if such a claim fails, it could tie up the GCC project in court for years and may even prevent them from distributing the compiler for a while and equally bad things. This way, they can point at the contributor as the guilty party, rip out the slop and replace it, and move on. The contributor may have a slight legal problem though...
Re: (Score:2)
Sovereign systems (Score:2)
https://www.scry.llc/2026/04/0... [scry.llc]
South Korea just pointed out the future...
https://www.scry.llc/2026/07/2... [scry.llc]
or you can throw up a digital.maginot line because you're so heavily ego invested in "code".
who argues over assembly code anymore?
Re: (Score:2)
AIs don’t know what their training material (Score:2)
Thanks, that makes the AI ban makes sense; they are afraid of AI generated code generating copyright infringement issues. The AI cannot claim it has never seen the code being infringed.
It absolutely can make such a claim when generating code based on its training . It does not know what sources it was trained on. But to be fair the far more likely scenario would probably be it searched the internet and found some relevant code to copy.
Re: (Score:3)
That's a question, but an even thornier question is whether LLM code is itself a copyright violation. Europe and the US appear to be moving, slowly, in opposite directions on this, so there's a risk that using LLM code will render your entire project copyrighted by someone else in various jurisdictions if they can show that the LLM you used was built using their code.
For example, CoPilot, IIRC, is trained on the entirety of Github. If someone finds their code on Github, they may co-own your CoPilot-authored
Re: Copyright (Score:2)
There are not enough lawyers in the world to handle cases like this. And most of these would be one-offs, so remedied in a class filing.
Remember SCO and Xenix? How quickly we forget. Expensive, protracted...successor cases are still ongoing.
Re: (Score:2)
Re: (Score:3)
For GCC it doesn't matter much if the code is public domain. Once linked to GPL code, you need to follow the GPL.
Before you object, that the GPL cannot be enforced for public domain code: You're right. But GCC contains enough GPL code, that the public domain code incorporated is hardly useful without using the GPL code along with it, and if you use GPL code, you need to follow the GPL. That's why GPL is compatible with licenses like MIT and BSDL, but considered 'viral'. It does not enforce the MIT part of t
Re: (Score:2)
Please read my post. It said exactly that: The public domain part remains in public domain. But in most cases it is not useful without the GPL part of the rest of the software, so any use will be covered by GPL.
I was wondering when FOSS projects would notice (Score:5, Interesting)
The problem is that LLM-generated code has no copyright. Anybody is free to copy it without any limits whatsoever. That does mean that the GPL can potentially go out the window if too much code is LLM-generated or LLM-derived.
And the second problem is that if the LLM stole too much from its training data, then the code retains the original copyright and ownership. Note that "too much" may not be a lot. There have been lawsuits for single lines of code (Java) or even interfaces (Linux Kernel) and these dragged on for years and did massive damage.
Due to these two reasons alone, using LLM code in any project that needs copyright (including all commercial software) is pure foolishness.
Re: (Score:3)
As a practical issue commercial software is almost always closed code, so nobody can see the source and simply search for anything that might share code with anything they have copyright on. The open source code isopen. So if you have anything copyrighted you can search the open source to see if anything looks like it is worth filing a lawsuit over. Especially if you have a competing product. So if you are say Intel you can check to see if any of gcc is similar to Intel’s commercial compiler and
Re: (Score:2)
Yes, the open source world really gets the short end of the stick in all sorts of way with GenAI.
You want to use an open source project but don't want to incur copyleft or even attribution? Launder it through CodeGen!
Your project has even two or three codegen fanatics? Congratulations, you now have inscrutable tickets and obnoxiously insistent pull requests that make no sense.
You have code unknowingly violating some closed source codebase? You can get hit out of nowhere, with no way of knowing you infringe
Re: I was wondering when FOSS projects would notic (Score:2)
Yeah but what are the odds that the training corpuses don't have tons of closed source code that just nobody has noticed is in there? While browsing the Web I've accidentally tripped over sources for SunOS and AmigaOS for example.
Re: (Score:2)
Yes. And while it will take longer, I expect we will see lawsuits based on object code as well. Those may or may not work, because object code only gives you an abstract view of the source code. But loops, decisions, call-graphs, etc. stay partially or fully intact. It may be enough to convince some court or at least drag on forever and cause problems.
Re: (Score:2)
Yes. And while it will take longer, I expect we will see lawsuits based on object code as well.
I’m a lot less sure. I would think a lawsuit could start based on object code, and during discovery get source code and continue if the source code looks similar enough. Doing a whole lawsuit on object code would face significant issues convincing a judge/jury that object code is “expressive enough” to manifest copyright. With the defense arguing that wholly independent implementations of something compile down into the same (or substantially similar) object code.
It should be “e
Re: (Score:2)
Re: (Score:2)
It's much easier to be a lawyer when you're allowed to just make stuff up like LLMs do.
Re: (Score:3)
The funny thing here is that the more serious problem happens when LLMs do not make stuff up...
Re: I was wondering when FOSS projects would notic (Score:2)
Some of it is also certainly closed source code that was leaked.
Re: (Score:3)
"So if LLM-generated code retains GPL copyright..."
LLM-generated code would be LLM-generated, it may be governed by copyright but it would not because any copyright were "retained". The code is newly generated, by definition.
Your assertion that "a lot of the training data is probably GPL code" doesn't mean anything. What would matter is if there is theft of copyrighted work. And LLMs may be able to use copyrighted works for training, just like humans can. It would seem to be fair use for LLMs if it is fa
Re: (Score:3)
The code is newly generated, by definition.
It is not. That you do not understand the law does not change the law. And courts generally do not buy all the bullshit about LLMs either. They see some data going into the machine and very similar data coming out and they will call that "copied" or "derived". An LLM is not a magic "code washer" at all, no matter what some idiots believe.
Re: I was wondering when FOSS projects would notic (Score:2)
"but, your honor, I ran all my drug money through the laundromat first to disguise the source of the revenue!"
Re: (Score:2)
"...if the LLM stole too much from its training data..."too much" may not be a lot..."
Any amount of stealing is too much, but LLMs cannot and do not steal in and of themselves even if they are tasked to steal. LLMs learn and then apply that learning, just as humans do. When humans do that, it is not called stealing. It is called ownership of what you've learned.
Current law is simply inadequate, there will no longer be a distinction between what is learned and what is copied when they can be essentially t
Re: (Score:2)
Bullshit. Code goes into box, very similar code comes out, infringement is achieved.
The LLM peddlers have known that all along, yet they still stole everything they could get their hands on. They did that because they know people like you are not smart enough to understand the situation.
Re: I was wondering when FOSS projects would notic (Score:2)
I think the test of copyright infringement is or should be simple. Is the code the same as mine. That should be the same whether the code was written by an LLM, by some code monkeys or by a human.
If people are insisting that the test should
Be whether code went into the LLM sausage machine, then they are not worried about copyright infringement.
Re: I was wondering when FOSS projects would notic (Score:2)
What LLMs do can not reasonably be called learning, or even frankly training, though the latter has unfortunately stuck.
On the other hand it's also not storing the original data (although the training corpus is that) but rather data about the data.
However, the standard for a derivative work is that it contains substantial recognizable elements, not that it's a 1:1 copy, so that distinction may be irrelevant, or at least not mean what you think it means.
Re: (Score:2)
Yes to both.
Learning involves insight and no serious legal assessment will ever accuse LLMs to have insight. As to derivative work, the only way to "clean" input is a clean-room process and that critically involves some person doing a creative process to turn a spec into a product. If that happens to create something similar, it is fine. But the key is the creative act that made the original is NOT copied and an new (not necessarily different, but necessarily independent) creative act is used instead. And a
Re:I was wondering when FOSS projects would notice (Score:4)
Yes, non-human code assuming that it does not itself violate any copyright does not earn copyright by itself. However, once combined with human output, the combined work is not exempted from copyright. So the 'no copyright' scenario would be no worse than pulling in code from public domain, the source may not be copyright, but the result after mixing it in would be. You also won't see companies even trying to argue that partial LLM content forfeits copyright for GPL, since the implications for proprietary code would be *far* worse given their near unanimous love of LLM code.
Your second point is the bigger gotcha, that an LLM generated copyright infringement can be a problem while every one actually involved can be genuinely oblivious. Even as an AI company has successfully defended themselves from being liable for infringing output, they essentially throw users under the bus by asserting that while infringement can happen and be facilitated by their models and services, it's the user's liability if that should happen and has nothing to do with their use of material for training and the resultant reality that the model *can* construct infringing copies of material in the training set.
Sadly, the second point is in practice waved away due to the assumption that the corpus of training material is almost entirely open source stuff, and while they do have copyright to respect, in practice they don't expect the open source projects to be resourced enough to detect and litigate infringement. In the "worst case" scenario, remedy is also likely to be easy because the courts will not recognize financial damages and the defendant could probably just LLM up a an alternative invocation, or 'shim' their way to isolate the code if copyleft, or simply credit the plaintiff if it's just BSD/MIT style. Long standing reality that violating open source copyright is "no big deal", only violating closed source copyright carries real risk. Some companies take the open source licenses seriously, but no company that has abused those licenses has faced anything vaguely significant in the way of consequences. I don't think anyone has been forced to copyleft their proprietary code due to a GPL violation by a court, at 'worst' they only have to correct the issue moving forward.
Re: (Score:2)
> The problem is that LLM-generated code has no copyright. Anybody is free to copy it without any limits whatsoever. That does mean that the GPL can potentially go out the window if too much code is LLM-generated or LLM-derived.
It is far too early to make this statement. It is sort of true in the US as a judge or two have ruled something like this, but copyright is a world wide issue, with a patchwork of completely different laws. It may not be protected by copyright in the US, but it might conceivably v
Re: (Score:2)
If you read the second paragraph of my posting, you will see that I have addressed that as well.
Re: I was wondering when FOSS projects would notic (Score:3)
These truly are uncharted territories so any overconfident claims about the legal status of LLM-generated anything should be viewed skeptically. Rarely is existing law well equipped to deal with new innovations, so it is doubtful that we will have any answers until new laws are written.
My bet would be that the tech oligarchs will write those laws, so they will probably seem contradictory in that they allow tech oligarchs to do whatever they want with data while placing onerous restrictions on end users.
Re: (Score:2)
Tech oligarchs do not write European law and they already are finding out that they are subject to it when trading in Europe.
Re: (Score:2)
It is intuitively NOT copyright infringement because when a human does exactly the same it is not. The law was clearly intended for it not to be.
Re: (Score:2)
Untrue on both claims. When a human reads code and then produces very similar code, that is infringement. You are _really_ clueless. The only thing that "cleans" code of copyright is when a human rites a specification (!) for code and then another (!) human does a reimplementation of that code based only (!) on that specification. That is called a clean-room implementation.
Re: (Score:2)
It is not, any more than when someone writes a book about elves, some Slashdotter parrots something they heard somewhere, or a half decent programmer writes any of the algorithms found in Numerical Methods. The GP is correct, copyright protects the specific expression, not the idea, and both law and precedent protect the case where you read something, put it down, then generate your own. Only substantial portions of exact text
Re: (Score:2)
And yet, there were long-running lawsuits about single lines of code or utterly trivial interfaces. Methinks you have no clue what you are talking about.
Re: (Score:2)
And yet, there were long-running lawsuits about single lines of code or utterly trivial interfaces. Methinks you have no clue what you are talking about.
Yes, there are long running complex lawsuits about all manner of things you might think are intuitively obvious, however just because the last lawsuit was long running and complex doesn't mean the next one about the same thing will be. In theory if the prior court found that header files are public interfaces and don’t have copyright protection the next lawsuit involving copyright on header files will have the defense lawyer point at the prior judgment and saying “we don’t need to litigat
Re: (Score:2)
So far the plaintiffs have tried going after the providers, and so far the providers have successfully claimed they aren't to blame, but have asserted that it can happen but that if it happens, it's the user who is liable.
So it hasn't come up yet, but it does seem like a company could come out of nowhere and hit you up for infringement because the model generated too-similar a copy. Despite not knowing, you still are on the hook, at least by the logic of the so far successful defense asserted by the AI com
Re: (Score:2)
Your claim has no merit. "No court has yet decided" is a non-argument. This situation merely means that the legal situation is unclear. Which is _worse_ in many regards.
Re: (Score:2)
You just overlook the tiny problem that parts of GCC could become commercial and hence distribution is not even possible anymore.
Re: (Score:2)
You know, ranting this way on the GPL does make you NOT look smart.
And they will know how? (Score:2)
Some are obviously AI-generated. But there as the code quality gets better how would it be distinguishable from human written? Less bugs?
And at that - would they use AI to find out if it was AI-written?
Re: (Score:2)
Some are obviously AI-generated. But there as the code quality gets better how would it be distinguishable from human written? Less bugs?
Just because code quality improves doesn’t mean the code looks like what a human writes. Currently AI code seems to not favor writing lots of functions, do it’ll basically expand out the same code in multiple places. Which I admit is also what jr programmers tend to do and maybe you can quality that as something that is code quality and will improve. You can frequently use a prompt to make it “don’t write the same code multiple times, define functions for anything repeated (maybe
Re: (Score:2)
Isn't it enough to go after people using em-dashes and people who can't draw hands? Next we need to claim people not using default arguments are robots?
Re: (Score:2)
Hmm - no my experience. Almost the opposite. Structuring the code is often hard for lazy humans. I see a lot of very well structured code from AI and this is more a telltale of AI. Good comments as well though their pedantic correctness is again purview of AI. (Mostly C++ and Python on my side).
But alas, by this flat refusal do they not shoot their feet off? There is a lot of good this tool can do. Rather, just penalize submitting AI slop as it burns reviewers energy?
Re: And they will know how? (Score:2)
Re: (Score:2)
In theory, same way they know the submitted code wasn't ripped of from the contributor's closed-source employer without authorization: they don't.
You can still have a policy to enforce, and in the cases where it is blatantly obvious, then it's easy.
In practice, nothing they can do to police the more subtle scenarios if the contributor flagrantly works to fool the policy. Except ban future contributions if they attempt to submit without disclosure and are caught.
Not banned but not allowed either (Score:5, Interesting)
This must be a difficult to stand for gnu to take at this moment, but I like its nuance.
Basically acknowledging that the code can be critiqued/reviewed by AI, and AI tools are helpful in discovery and research, but that any substantial (>15 lines) contribution of code needs to be human-created.
I think we're at a strange crossroads in AI. Many people are willing to accept the downsides of it (intellectual property, energy resources, ownership by companies that are unlikely to consider your welfare as their priority, privacy issues, downstream cost issues when profit is needed, dependence, negative impact on our own cognition and others) for what are now clear benefits.
There have been many times in past humankind where populations have chosen what makes the day-to-day easier or more productive, despite it clearly being wrong - is AI one of these situations?
Re: (Score:3)
This must be a difficult to stand for gnu to take at this moment, but I like its nuance.
I find it interesting that their intention is clearly based on concerns about IP ownership, unrelated to quality. They're not banning "AI slop", they're banning code with unclear ownership. They clearly want to avoid a situation where they take a code contribution which the AI model owner might claim to own, or which owners of the AI's training material might claim to own. If those claims were upheld, it would mean the contributor didn't own the copyright and therefore couldn't license it under the relev
Re: (Score:2)
Torvalds said it already more clear that if you don't see that AI is becoming better you're a fool.
If you're wanting to ban AI for your personal dislike but claim you do it because of quality concerns, your standing will be weak when models are becoming better and better. So your only option is to avoid the strawmen and tell what you dislike. In case of GNU it's the copyright concern, so they say that and can keep their policy until enough courts have decided on that.
Re: (Score:2)
> There have been many times in past humankind where populations have chosen what makes the day-to-day easier or more productive, despite it clearly being wrong - is AI one of these situations?
I asked Google AI, and this was its response:
Historically, humanity has frequently embraced convenient or productive technologies -- such as fossil fuels, industrial agriculture, and single-use plastics -- only to face severe long-term environmental, social, or ethical consequences. Whether artificial intelligence
Good (Score:5, Interesting)
Re: (Score:2)
Re: (Score:2)
Modern software increasingly does not depend on traditional compilers. How terrified will you be when there is no compiler at all?
"This is the tool you rely on to make your code."
Tell that to Python users. Do not worry though, there aren't many of those!
"...then that's just being naive."
Naive is thinking this problem hasn't existed for decades. It's been well understood before you even went to school. It's NOT an AI issue.
"given enough eyeballs, all bugs are shallow" (Score:2)
Re: (Score:3)
AI "anything" is worse than human equivalent, AI has no values. It's faster, so what? It's easy to go fast when it can be wrong.
The overwhelming majority of features that go into making a human mind what it is do NOT have analogs in LLMs nor do AI companies have any interest in making that happen. Instead, AI is a billionaires' rush to own the one interesting thing that has recently developed. Claiming that AI is about to produce super-human intelligence is really a monstrous insult to everyone's intell
Re: (Score:3)
Re: "given enough eyeballs, all bugs are shallow" (Score:2)
Re: (Score:2)
Code analysis is a bit more unambiguous in terms of value, as long as the findings aren't taken *too* seriously. It has impossible attention span and can spot real issues, or failing that a mistaken issue can highlight something 'off' and inspire a true finding. It will also indicate lots of false positives, but at least the consequences can be limited to wasting your time so long as you take care to analyze.
The code gen is a bit more harrowing since it requires human vigilance to correct the problems in
Re: (Score:2)
Re: (Score:2)
Yeah, LLM is tricky since sounding correct is the priority, and being correct is a side effect of sounding correct.
Business executive style behavior.
Re: (Score:2)
Legal issues (Score:2)
A lot of comments are talking about the legal pitfalls, copyright, licensing, etc.
While these are important issues, GCC can't be serious about not allowing people to use AI generated code.
No one doing anything complex with software is writing their own code anymore. (See Andre Karpathy)
Re: (Score:2)
> No one doing anything complex with software is writing their own code anymore
You are not everyone.
The fact you're lazy and do not care about the quality of your code or its future maintainability doesn't not mean anything unless you're reflective of the entire industry. And you aren't.
Re: (Score:2)
Not yet. But I won't try to become a no-AI nomad, if I were you. Look into social media to see how AI critics have one meltdown after another when yet another project adopts AI. Developers like powerful tools and as soon as AI becomes useful for them (I guess everyone has an own level how powerful a model needs to be so they can use it efficiently) they use it. Yeah, some may object for personal reasons and some projects may more or less ban it, but as a user you will have to deal with programs containing A
will not end well (Score:2)
This makes sense now due to legal issues. But it will soon be the case that AI-generated code has fewer bugs than human-generated code, and code review by AI will be more accurate than code review by humans. At that point, projects that prohibit AI will fall behind.
Maybe code really shouldn't be copyrighted at all. Rather maybe algorithms should be patentable.
Ultimate plagiarism (Score:3)
Correct me if I'm wrong, but isn't AI is just one, big, ultimate plagiarism engine?
The creators of AI models are so aware of this that some of them actually added config switches to hide code attributions in generated content...but without stopping models from using copyrighted code in the first place. Just hide attributions and play stupid.
Re: (Score:2)
Yes you are wrong, at least for how you phrased it. There are no config switches to hide code attribution. Most inference software is open source and you can search for them. Professional hosters often use vLLM, private users llama.cpp. Both repos are open for you to search for the "hide attribution" switches. Models are also not using code, as they don't have access to their training data at inference time. If you want to debate copyright, you need to target the training process and if the model trainers w
Re: Ultimate plagiarism (Score:2)
The training process is not everything. One example is that models, after being trained, have external interfaces to fetch outside data for use with the model, augmenting the model's own data store. That's why, for example, they can answer questions about events and data being created at the moment. So even if you target the training process and make sure that you train on open data only, which is near impossible in practice, you are still wide open to infringing while using the model.
And no - it's you who
Re: (Score:2)
The burden of proof for the switches is on you.
And you're moving the goalpost when you say "I can use a harness that fetches code from somewhere else"
Re: Ultimate plagiarism (Score:2)
There you go. One of many examples:
https://docs.aws.amazon.com/am... [amazon.com]
Turns off attribution but doesn't stop the model from using the code.
Re: (Score:2)
The feature is exactly the opposite: Other than other tools, the Amazon tool has a feature that tries to find if code is similar to existing code. This feature can be turned off. So the difference is here, that the feature tries to find similar code at all, even though you usually don't need to.
GCC is following GNU's lead (Score:2)
The GCC policy is largely in alignment with the greater GNU project policy (under which they reside).
The more interesting question will be if individual contributors (many of which are employed by major organizations that may already be using various AI enhanced tooling) need to change their development practices (and will they).
Lie to me! (Score:3)
All this will accomplish is to encourage contributors to lie. Or not contribute.
You cannot put the genie back in the bottle.