The issue is not and has never been “AI bad.” The issue is that everyone is trying to use it as a replacement for human generated content and human critical thinking. If you don’t know how to math, how could you possibly know it’s wrong when a calculator tells you 1+1=11?
Ok, but the AI actually is bad. Like, it’s bad at writing code. The code I reviewed was sloppy at best. If I got that from a high school student for an assignment, I would give it a B. If I got that from a college senior, I would give it a D. If I got that from a junior dev, I would give them a serious lecture about testing their code.
Right, but AI in itself is not bad. Instances of it are bad because it does certain tasks poorly, and when we don’t know our topics well enough, we can’t spot its errors. You don’t learn when you expect your shortcut to do the work. And worse, you can’t error check if you can’t do yourself, what you’re asking it to do.
The right answer is obviously 10.
Human bugs tend to follow patterns, quite consistent ones for the same person.
Further when searching for bugs code done by the same person is quite consistent so somebody else looking through it will figure out the pattern an much more easily navigate it.
Last but not least, the human actually learns and will stop making certain kinds of bugs, especially if they’re the one having to find them out and fix them.
AI bugs don’t follow patterns, the code isn’t consistent across the code base and it can’t learn.
I get the impression that this Lutris Dev isn’t even at the expertise level of Senior Developer.
PS: Oh, I forgot what’s maybe the most important thing - humans try a lot harder to avoid making bugs which can cause huge problems or at least will put measures in place to avoid huge consequences if such a bug exists (the probability distribution on the severity space for bugs made by humans is not uniform but rather weighed towards bugs with fewer consequences), whilst AIs do not, being just as likely create bugs with massive impact as cosmetic ones and not really creating “just in case” checks to catch the nastier bugs.
My house already has roaches, might as well let the termites in too
I started giving Heroic more consideration. And holyshit it is much better!
I started to use faugus launcher instead.
I do wish the gaming meta package in Catchy Hello allowed to only remove Lutris
Are we gonna post everything he say here now? Just let him be at this point
At this point, I’m assuming OP wasn’t on Lemmy last week.
(Possibly wrong, but it’s a plausible explanation of why this kind of thing happens all the time here.)
Last week, he was saying that everyone was a bullshitter because no one could point to any low quality code the AI produced. So, I reviewed his commits and of the four I reviewed, two had bugs.
Thanks for the bug reports!! Quite honestly, if that’s the level of bugs we’re dealing with due to our use of AI tools, that’s a pretty good deal. I’ve seen much worse, so much worse in code that we actually shipped in releases. And no AI was used to create those critical bugs. Now we’re dealing with edge cases. Like that whole extractor bug for AppImages. This was introduced during the development of the GameJolt service which can have a few different ways to package games. The key here is to have a very solid and complete test plan and minor bugs like that will get caught.
This is absolutely not the “technical debt” and “bloat” and “vulnerable code” I’ve been hearing about. Those are bugs that aren’t nearly as bad as the ones we’ve been used to over the year. And they’ll manage to get fixed like every other bug.
The full comment, nowhere does it day he’s “fine” with AI generated bugs, only that the frequency of them is comparable with non-AI code.
(Not that I support his actions, ofc)
“that’s a pretty good deal” is him saying he’s fine with those bugs.
He’s also heavily downplaying the severity of that bug. If a user hit that bug, it would keep copying that AppImage file over and over until it filled their disk and crashed the app. Then the user would have to figure out what happened and where all those duplicates were to fix their system, all while things were falling apart because nothing could write to the disk.
Many systems cannot successfully boot if the disk is full, so those users would probably have to reinstall their system if it crashed or they rebooted, and they didn’t know how to navigate a root shell. Even if the system didn’t crash, many apps won’t start if the disk is full, so the user is just going to have a really bad time overall.
Later in the thread, another user defended the severity of the bug by pointing to other bugs that Lutris has shipped which have damaged their users’ systems.
It’s also worth noting that I only reviewed four of his commits and found two bugs, one severe. So the frequency of these bugs seems much higher than without AI tools. Who knows how many others the AI has introduced, but I’m not going to review all of his slop if he can’t even be bothered to do it properly before he commits it.
brokenglasszero
last week
It’s one thing to decide to use AI tools and such for the work you’re doing. You can do that, and I don’t have to like it and both are fine. The real painful thing about this is “Anyway, I was suspecting that this “issue” might come up so I’ve removed the Claude co-authorship from the commits a few days ago. So good luck figuring out what’s generated and what is not.”Because you’ve decided to do this, we can’t even fork it and we have no idea if we’re even adhering to the license properly, no idea where the code came from, etc… You’ve basically poisoned the Lutris source code to spite everyone and it feels incredibly against the entire spirit of what open source should be. It’s frustrating to seeing a project I really enjoyed taking a turn like this, far more so than the usage of AI alone.
He readded the attribution last week -
https://github.com/lutris/lutris/discussions/6530#discussioncomment-16107836
If he can’t be bothered with the project anymore, just give it up, don’t resort to shitty ai.
Sadly, it seems to be fairly common to have at least some AI slop code now. E.g. lemmy itself appears to be planning to do so too.
It’s like having slop would get you some prize.
I think you misrepresented it. Lemmy dev says it needs to be declared and AI is useful for some operations but it must pass review.
I don’t use AI, and think the code is crap, but assisted is different when in skilled hands.
The only issue here is your absolute no vs lemmy’s pragmatic no unless used as a tool in small instances. There are better hills to die on than this and you’ll just lose support for whatever objective you have.
There are lots of legal problems with accepting any AI generated code, regardless of whether it’s bad quality or not. For one, the AI tends to reproduce copyrighted code without a proper license:
https://youtu.be/xvuiSgXfqc4?t=247
Another is that AI generated code is not copyrightable, so even if it’s not copying someone else’s code, it can’t be licensed under an open source license.
How would such limited use fix the plagiarism? Here’s a lawyer demo’ing the issue: https://github.com/mastodon/mastodon/issues/38072#issuecomment-4105681567
This isn’t a legal advice. Check out the link, form your own opinion.
I watched. He used a prompt with exact wording from an example and obviously it’s the most logical continuation, so AI would generated it. We know how they’re trained. But how many open source prompts will start with exact code and comment? Unlikely to ever happen in real world. So unlikely to he direct infringement. Someone could easily sue the AI companies with these examples to prove it infringes copywrite work, but then govs are going to protect them so it won’t happen.
It’s unlikely to get identical output without intention, but it’ll also take the infringed actively taking steps to sue that case. Stastical likelihood of this happening in real world is low.
I’m with you in wanting to watch the AI industry collapse. But I’m unfortunately in a minority without the lobbying power, so it won’t happen.
So is
function isEven()a prompt with exact wording from an example, too?
define slop
thats a real definition but the person i replied to, wasnt using it in that way.
It’s easier to troll if you wait until your account is over a month old ;)
i am not trolling. it was a stupid comment
This is bait.
I think Ricky will switch to Bottles.
I wonder if Bazzite will stop shipping Lutris because of this
They are thinking about this for a while now, in January they announced: “Lastly and on an unrelated note, we’ve also begun testing Faugus Launcher as a potential replacement for Lutris in our :testing branch. If we move forward with this change, we’ll provide at least six months of advance notice before removing Lutris so you’ll have plenty of time to migrate your prefixes.”
I never really used Lutris, but I liked what it was for so I never removed it, but when the news broke I went and took the 3 seconds to uninstall it
Edit: ok this shows I’m not a super adv user or anything, but I thought it was removed just by uninstalling from flathub. Embarrassing.
Last time I’ve tried Bazzite, Lutris was installed as a system package, so you can’t easily uninstall it without rebasing the system image
Personally I didn’t like Lutris’ reliance on Proton and the Steam runtimes, at this point I could just use Steam directly as my launcher for third-party titles
Personally I didn’t like Lutris’ reliance on Proton and the Steam runtimes
What exactly are the alternatives? Wine isn’t optimized for gaming and Wine-GE isn’t being developed anymore in favor of Proton-GE.
at this point I could just use Steam directly as my launcher for third-party titles
Lutris is much more light-weight than Steam, which is basically just six chrome processes at this point.
And how do you do that? Everything I saw said “Don’t, it’s too difficult. And if you do, your system will just run worse than if you ignore that it exists.”
I’ve enjoyed Bazzite in my first few months off of Windows, but between this and a few minor annoyances with just how immutable an immutable distro is, I’m having second thoughts on my choice of distro.
I would kind of like to be able to uninstall it from my Bazzite box.
They better not… Its far to fundamental to the system to get eid of it over proclivities about AI.
If he really means it like this, then this person is no longer trustworthy for me.
Is lutris even used anymore? I would assume with proton being as good as it is now, lutris is kinda useless.
You use lutris with proton.
According to Flathub it gets over 75k downloads a month.
Crazy ngl.
I’m glad you didn’t lie about thinking that was crazy!
Lutris is like heroic or steam: it’s essentially the downloader and launcher for games that are then run by proton.
That’s not what he said.
It’s paraphrasing, but that’s pretty much what he said:
Quite honestly, if that’s the level of bugs we’re dealing with due to our use of AI tools, that’s a pretty good deal. I’ve seen much worse, so much worse in code that we actually shipped in releases. And no AI was used to create those critical bugs.
Libel requires the statement to be false.
Libel requires the statement to be false.
Indeed. And your statement that you now describe as “paraphrasing” was false.
In what way is what I said false? His statement describes the bugs as a “pretty good deal” because he’s seen “so much worse” in his code without the inclusion of AI. Therefore, he’s cool with AI generated bugs because his code is already full of bugs.
It is false because it is not what he said, and does not have the same meaning as what he said.
Moreover, your misrepresentation is damaging to the developer’s reputation, and misleading to everyone reading here. Please stop.
In what way is what I said false?
Do you think he’s not cool with AI generated bugs in Lutris? Do you think the code isn’t full of bugs? Do you think the reason he’s cool with AI generated bugs isn’t because his code is already full of bugs?
It certainly seems like all of those elements are in what he said. He knows that the AI is introducing bugs (I pointed out two bugs that it introduced in that thread), and he’s fine with it (he said it’s a pretty good deal), because the code base was already buggy before (he’s seen so much worse in code he’s shipped in the project).
He kept challenging everyone in that thread to find below average code pushed recently. I took him up on it, and looked through his last four commits (all attributed to Claude) and found two bugs. He is totally fine with that. If it were me, I would really rethink using a tool that introduces bugs in half of its commits.
people are asking how exactly it is not/does not have the same meaning as what he said
Considering how often and how heated the topic comes up in lemmy (even though the actual discussion takes place at GitHub) this is some sort of bullying.
Instead of simply parting ways some are harassing the developers of a free software in order to gain exactly nothing.
I’m involved in that discussion because I like Lutris and don’t want the project to suffer because of the use of AI tools. The developer challenged people in that discussion (myself included) to find low quality code that had been pushed recently from the AI. I did. Two of his last four commits introduced bugs.
I know that you are involved.
You ain’t a dev or a maintainer oof the project, so keep the discussion civil. Also stop trying to rally people for your “cause”. At this point, you are just bluntly trying to make someone miserable.
There is nothing to gain for you.
Holy shit you have AI derangement syndrome. If someone even mentions AI, then your instant reaction is to lie about it.
deleted by creator
These people are ridiculous. This is an open source side project, if people don’t want “ai slop” in their app, then maybe they should fork the project and maintain it themselves.
They can’t fork it now because the fact that the AI commits aren’t labeled as such means there’s no way to tell which need to be removed.
They can. They just need to find the first instance of AI, branch from the commit before and start with that as a basis for master. Then push that branch to their own remote.
If someone cannot do that, they probably aren’t competent enough to maintain a project of this importance. They can then cherry pick commits that are good and merge those. Or request others recreate new PRS with them with correct attribution. It’s tedious, but easy.
Maintainership isn’t fun. The hardest problem is finding someone that cares enough to take it on. Many would give advice if someone was willing.
Sounds like Lutris dev was burnt out, so it would probably quickly replace it.
They can’t fork it now because the fact that the AI commits aren’t labeled as such means there’s no way to tell which need to be removed.
So… they can’t do that because they can’t tell the difference between the human code and the AI’s code? So that means that either A. The human code is also slop or B. The AI’s code is on par with the human’s code. This comment really proves that this aversion to AI is purely ideological.
Being able to see a difference in code quality is one thing; being able to prove who wrote the code for purposes like license compliance is another.
Based on the thread I originally linked, and the dev’s response, with regard to Lutris, I think the answer is A.
But there was attribution, so you can easily check out commit before attribution was removed, use git blame on class attribution to see when it started, and fork before it.
What I’m saying is it’s probably not worth forking Lutris because it’s bad code. It would be better to just switch to a better alternative.
Like?
Heroic doesn’t work in all cases. I have to have both installed to use my library.
I’ve had many issues with other software and ended up having to use Lutris.
Lutris may have bad code, but functionally mature is generally what succeeds. Sexy code that doesn’t work exists on a handful of computers only.
There is a reason we are talking about it. Its the most popular Linux launcher. Wishing that wasn’t true doesn’t change the fact it is.
