Since we last covered the way that the Linux Kernel project has been dealing with the influx of code created using LLM-based coding assistants, all the way back in April, things have changed a lot, with most recently the Linux maintainers deciding to fully drop the requirement to mark LLM-assisted commits with the Assisted-by tag.
While not a fully uncontroversial decision – as also reported by Phoronix – with key maintainers already having made this decision it seems certain to proceed. This is not the only change made fairly recently, with a ‘Sashiko’ bot introduced that is supposed to handle code reviews.
In a July LKML post [Linus Torvalds] defends this decision, with a follow-up post in August by [Jakub Kicinski] it being stated that budget has been secured for using Meta’s LLMs, using said LLM-based resources for reviewing patches, as well as creating and editing commit messages. As noted by [Jakub] they’re running multiple frontier models on each patch to try and weed out confabulations.
Whether this will prove to be the right decision remains to be seen. Whereas most open source projects have firmly banished or heavily curtailed LLM-based contributions on slop, copyright as well as licensing grounds, the Linux Kernel project may be an interesting experiment in what the outcome of the exact opposite approach will be.
Someday there will be retrocomputing articles about people going back and programming in BASIC or whatnot “the old-fashioned way,” where you have to memorize syntax and commands and stuff. People will do it for nostalgia’s sake.
Like my grandpa punching tapes near the fireplace, drinking Five Alive, sure.
That someday has been the past decade or so, it’s got little to do with LLMs.
That said, maybe we’re headed to that scene in Back To The Future 2 after all – “You have to use your hands? It’s like a baby’s toy!”
If the companies manage to stuff real logic processing in their talking-machines, we’ve got other problems
Or maybe they will do it because they’re being fascinated by building something from ground up all alone?
Like model makers, makers or radio amateurs do?
It’s an surprisingly satisfying experience to create or accomplish something on our own.
Like learning a new language, building a bird house or a model railroad.
In fact, many of the young are being fascinated by old tech because it’s a totally new, so very different experience to them.
Be it VHS, a rotary phone or a typewriter or old computer. They love examining or using that stuff, even.
Because they’re tired of tapping on a glass plate (smartphone) since they were born.
They want to feel something, want being challenged and want to learn to fix things.
Here’s a video I’ve found that mentions this: https://www.youtube.com/watch?v=0dEJiQnotR8
Feels like it was a contractual decision and that it was more of a compromise rather than a desire.
How so? What was the contract that was being decided, and what was the compromise a compromise of?
Feels like an irrational take to protect your own interests.
It’s really sad to see even the smartest people throwing more and more money into big tech pockets. No one can run these frontier models on their own independently, they have manufactured a need, and everyone is becoming more and more dependent on them. It’s so depressing.
They’ll democratize because of two factors:
– optimization of models and specialty models
– cheaper and ever more powerful mass-produced AI chips
The advantage of big AI companies will not stay for long. One does not need a cutting-edge, god-powerful model for programming task. “Good enough” models will even run on the edge soon.
No one needs models at all for anything. Let alone the buggy slop they generate thay is fully of security bugd.
I’m glad you are smarter than Linus and the other people working on the Linux kernel.
You should them! Also tell them to remove Sashiko.
Smart persons can also be wrong.
Why do people think that person is? Is there any proof? IQ/EQ test etc?
Being successful and being smart are separate things, I’d say.
Let’s just have a look which kind of people do lead companies and nations..
The buggy slop that built an entire big free app for me last week?
You literally do not know what you’re talking about.
Or you will learn about it soon. Also, will you take full responsibility for the LLMs app?
Link(s), or it it never happened.
I’m glad we have geniuses like yourself to help wrote code thay is fully without security bugd.
You’re a couple years behind the curve. It’s ok. Time passes everyone by eventually.
Also, 99.9999999% of all security bugs come from humans. Due to AI, we’re discovering and patching more and more bugs than ever before.
AI invents novel zero day exploits to hack websites it thinks will help it cheat at a benchmark… and you think it’s somehow BAD at security?
Inventing facts and confusing things is literally one thing LLM are good at.
You can download models for free and run them on your gaming PC. Paying a Netflix like fee for ChatGPT and getting a really useful tool instead of a bunch of time wasting television doesn’t seem so bad to me. It’s all about convenience. If you don’t have a GPU, get one or pay the small monthly fee for ChatGPT/Gemini/Claude/Whatever. Or just keep doing your hobby the way you always have. Or change hobbies. No one is forcing you to do anything. You shouldn’t be depressed because other people are making decisions for themselves and not forcing anything on you.
People are becoming dependent on LLMs like people became dependent on computers. You’re not required to use either to do woodworking, but woodworking can be done with the help of computers. It’s a personal choice. Do what you like and stop worrying about what others are doing.
As someone who grew up along computers, picked it up as a hobby first, and eventually made a living and a career by working in IT, yes, AI is very much being forced on me. I try to learn what I can, and keep doing things as I always have where I can, while I watch helplessly how my job, my hobbies, everything gets worse every day as a direct result of AI getting shoved into every aspect of our existence. How could I not be depressed?
If it doesn’t seem so bad to you, I really wish I could see this reality through your eyes. It’s all about convenience… whose? Even you are telling me to buy their GPUs (which, like most hardware, is no longer affordable now), or to pay their subscriptions. For what? Just to do even more work for them?
AI could be a very useful tool, but wielded by corporate greed it’s just high-octane fuel for enshittification.
Nonsense. You can run many of the free and open models on you own hardware. And they get better and better. Nobody is complaining we got dependent on cars. And that we should all user our legs. Or that your cooker is to convienent and you should roast over a wood fire.
Linus is right. Its a tool. I now spend more time thinking actual architecture and functionality that I spend tweaking and fixing code and anoying bugs. And fighting unclear documentation. In short. More time for my actual job, which is not described with typing.
I don’t understand the purpose of this. Wouldn’t the Assisted-by tag at least be useful for bookkeeping purposes?
I’m with you. It seems to be simple to add and would help in keeping track of things.
Even if all it did was make it easy to automate using other AIs, that weren’t part of the tag, to check the code like getting other people to look at your code.
In general, it seems like a tag that is only useful.
While true, and I even agree,it won’t help.
People lie. AntiAI people will not add it as they start to use its advantage and start using it.
AntiAI people will moan and cry and be troublessome whenever they see the assisted by tag.
It has become politically loaded and to what purpose.
I’d imagine in the next 5 years all written code will have some form of AI assistance.
Tab compleyion? AI instead of google searches? How will you be able to avoid it then?
To me, AI has proven its worth, and for a toddler it’s doing fine already. If we double the capabilities in the next 5 years and after that it falls flat and is at it’s best, thats already huge.
For the doubters, just look at all the found security issues today. If that’s not a win, idk.
As Torbals put it, natural intelligence isn’t all that amazing all the time either, as per above paragraph.
Letting AI run my itself? I doubt that. For now, it can’t thinknso doesn’t know intent, the future and thus not what good architecture or good design is.
“If we double the capabilities in the next 5 years and after that it falls flat”
Then what would we do, when that model produces a bug that even it can’t solve? We as human will be powerless, because we don’t understand what the heck did it write in the last 5 years.
He didn’t say that we should also turn our brains off; in fact, he explicitly said that turning it loose by itself wouldn’t work out generally.
To deny any utility it provides is purposely blinding yourself. There have already been clear demonstration of its usefulness.
True. But a problem here is the flawed (US) corporate logic (which is being exported to whole world).
It’s all about competition and maximizing of profits or efficiency without any regrets.
Which means that everyone is trying to beat the other one, at everything.
This will lead to an increased downwards spiral in which the human factor is becoming less and less valuable.
Production quality drops, loans drop too, employees (being the consumers same time) have less money to buy, economy breaks down. Finito.
I mean, let’s just think about it.
It’s as if some pupil was starting to cheat on a math test and gets away with it.
Suddenly, all the other pupils who notice it will try to do the same.
In the end, most of class may passes the test but has learned none.
The teacher or school district etc then may figure out the math tests were too simple
and rise the bar in next year, making it more difficult.
Suddenly, all the people on that school or in the area are required to rely on cheating because they would nolonger pass the math tests without it.
This will cause the honest, upright pupils to be left behind.
And that’s working against of the original purpose of the whole thing (the test, the tool).
This is sort of like a domino effect, in which a tool (the cheat) becomes a trap.
It makes people depend on it, because of the silly concept of being competitive no matter what.
That’s the danger of A.I., I think. That we might end up being helpless without it.
Just like some people are lost without a smartphone already.
They may rely on Google Maps or a Navi to get from A to B.
They can’t even use a paper map anymore or might dependend on a language translator app,
because they live in a foreign country they haven’t bothered to learn the language yet.
I mean, many can’t even read an analog clock face anymore.
They are unable to use any traditional mechanical clocks, thus.
Such as the ones found on public buildings (town hall clock).
And I’m not even mentioning the lacking ability of reading coursive writing in a historical context. 🥲
PS: My apologies for my poor English. I’m not a native speaker.
Cool! With that uncritical thinking of people we’re being lead straight into a dystopian age in which “engineers” nolonger understand the technology they’re using.
It’s causing the reversal of control, in which machines control us.
Machines that understand our “functioning” and habits better than we do theirs.
It’s the beginning of giving up of self-responsibility.
People nolonger have to grow, learn or be responsible.
Who needs skills anymore if they have “A.I.”?
While it may sound logical that A.I./LLM be helpful, it misses one point.
It requires people (users) who are mindful and responsible.
In the hands of the unqualified or lazy, A.I./LLM won’t be an aid.
It will simply do the thinking for them, causing further degeneration of society.
Let’s just look at the tech-savyness of the smartphone generation.
It’s a generation of consumers, rather than users.
They use the cloud and give their whole personal life out into the hands of big companies.
They already lost control about their lives and don’t even realize it.
By contrast, in the 80s and 90s, the average users had some computer skills.
Many bought extra documentation (books, magazines etc) to deepen their understanding, even!
Yes, there were DAUs, too who used whiteout on screen or broke the coffe-cup holder.
Such users always existed. But the average users had an basic understanding about technology, at least.
Thus, it might been a consideration to limit A.I./LLM access to people who are mentally mature enough to use it with care.
People in science or research. Or other institutions, such as art schools, not sure.
Or who have passed a test that certifies that they do understand A.I.
Let’s think of it like aquiring a driving license or a computer driving license (ex ECDL etc).
Letting people use A.I./LLM for fun is one thing, but letting them use that for everything else in life is at least questionable.
Using it for work/business could be seen as a form of fraud.
Because someone else is doing the actual work (the A.I./LLM).
The removal of the tag is nothing more than gutlessness, I think.
The decision was made to avoid any social backlash. They’ve chosen the easy way out.
You’re reading too much dystopian science fiction and confusing it with reality.
You don’t even have good material for your generational mud slinging (which you should avoid; it’s just evidence of a low effort thought). The tech savviness of every generation sucks. There’s always a handful of people interested in it.
The “average user” in the 80s was who exactly? Who was buying home computers in the 80s? Certainly not the average person. And that’s your fundamental problem. You’re not capable of distinguishing the difference between niche hobbyist usage and widespread usage. If everyone had a computer in the 80s or 90s, the vast, vast, vast majority of people wouldn’t be tech savvy.
It was a hobby that technical people got into, that skewed results and you can’t get beyond that bias it seems.
The rest of your technological ranting is just as misguided. And that’s all the reason to remove it. A bunch of misguided people on the internet being angry at something they don’t understand is a dangerous thing.
@Frink Hi there! Personally, I think that keeping discussing such things itself might avoid a dystopian society.
Perhaps its not even so important who’s right or wrong, as long people do retain critical thinking and remain capable of cognitive self-reflection.
Because nothing is worse than indifference.
About the old story about being able to afford a computer or not: Good point. Maybe I’m wrong, that’s fine.
But how to get an definitive answer to something that has no records/numbers being documented?
I mean, owning isn’t even a criteria, maybe.
There were and are many kids that have no game console but play video games at a friend’s house or at the arcades.
Same goes for computers. There were/are people who use(d) them at workplace or at a friend’s house or relative’s home.
Personally, I just noticed that many people I talked to in my country/region/social circle did fondly remember using a C64.
Many weren’t even into computing hobby but were ordinary people.
They had them at school in the computer room (-or computer lab, IT class, whatever-) or used the C64 in the children’s bedroom as a typewriter etc.
Such things. Same goes for DOS PCs or 90s era PCs in general.
I remember ordinary people mentioning Norton Commander or making jokes about doing a “FORMAT C:” to clean up a PC. Such things.
The memory managment and driver installation alone was a whole story on its own (MemMaker, autoexec.bat, config.sys).
Ordinary people were not excepted from this! They had to figure it out, simply.
I’m not trying to say everyone had owned a PC or home computer, but that people who had to get along with one had a learning curve.
They had to invest mental work to operate it, buy a book, enter a course etc.
Even if it was just an application specific thing (WordStar, Lotus 1-2-3, MS Office, MS Works etc).
And what they learned often stuck with them, which was good.
Commonsense, a rich vocabulary and some general knowledge don’t hurt.
The world might have a need for autodidacts more than it has for A.I., who knows? 🙂
Any AI code in the kernel breaks the GPL license. Since AI code is inherently not GPL compatanle at all.
I’m glad you are smarter than Linus and the other people working on the Linux kernel. I’m so sure they haven’t thought of this!
Go be condescending somewhere else. You’re annoying.
Go be ignorant and knee-jerk reactionary somewhere else. We like to debate, learn, and try new things around here.
Since they are allowing AI code in the kernel they have apparently not. AI code is non-GPL and also based off stolen work of various incompatable licenses.
Once you say the work was stolen, you defeated yourself.
Please spare us the “AI just regurgitates other people’s stolen code” chestnut. It’s wrong. A human programmer had to read countless other pieces of source code as well to learn the skill, that doesn’t give the original authors ownership of the programmer’s own output.
I see. There will always be people who see absolutely nothing wrong with a person who spends a lot of time tracing Scooby Doo characters and then goes around drawing Scooby Doo characters with an artist’s pride.
Focusing on the act of learning or rote memorization is a red herring. The thing we need to protect at all costs is humans knowing how to teach others.
Nothing is going to change.
The entire chess world is watching LLMs unfold and the underwear on head hysteria people have over it, all the while thinking “over reacting idiots, it doesn’t actually change anything for the worse.”
+1
If they are no longer tracing and are creating new works then they should have an artist’s pride.
I’ve listened to a lot of artists talk about how they started out by tracing other art and using it to learn the rules for good art and design.
AI code is purely code from non-GPL compatable sources
Kernel maintains this: “ You are expected to understand and to be able to defend everything you submit. ”
Maybe you can be a crappy programmer, but you’ll still have to be one.
“Intellectual Property” requires intellect.
What would that bookkeeping be based on? What are you keeping track of? If the code is good, who cares whether it was made by hand or by tool? The name of the person who submitted the code is still there. They don’t tag which IDE the person used to write the code, nor the plugins they have installed, nor whether they made the code using a Mac, Windows, or Linux. They don’t want to keep tagging unimportant information. Seems logical to me.
Yeah, let’s just use any source code and sign it with our own name! 😎👍
Just like young Bill did when he searched the paper wastekaskets of co-workers for good code snippets in the 1970s or so. Yay! 🥳
Seriously, though, what was the original idea of the GPL when it came to life? 🙁
Wasn’t it about giving credit to the rightful author
and protect him/her and his/her work, while simultanously allow others to participate on the software and learn from it?
What has GPL, GNU and Linux has become? Is it still true to itself?
Or are we gotten old enough to watch the hero(s) to become the villain(s)? 😟
It would let them keep the genetic diversity of the code healthier by running it through other models besides the same one or people can learn the negative quirks associated with models and either improve them or at least watch out for the same errors in new code.
It’s meaningless.
Define “Assisted-by”
The BDFL is now 56. At best he has maybe 10 more years of productivity left in him.
Once he’s out of the picture, Linux is likely to become a design by committee shitshow.
This.
But the question is whether AI will be able to ruin the kernel fast enough, before gen-z will rewrite it in rust?
gen-z is capable of coding ? that’s a bold claim !
Hi there, sure you mean “gen z” and not “gen alpha” ? ;)
https://m.youtube.com/shorts/fkf4myBkHDw
I mean, the older members of Gen Z are born in the 90s.. The 1990s!
And VB 6, VB.NET and Delphi 7 were popular up to mid 2000s.
It wasn’t before 2006 that good old Windows 98SE had retired.
The “OS” that still required MS-DOS knowledge to fix config.sys, autoexec.bat and co.
So if a Gen Z was born in, say, 1997 it was almost age 10 in 2006!
10! A whole decade! Tech savy kids started programming way earlier sometimes.
I was about age 6/7 when I programmed in QB, VB and Turbo Pascal.
But that was nothing compared to the kids of “Generation C64” who grew up with BASIC driven home computers much earlier. :)
These kids typed down listings from magazines or altered the BASIC programs etc.
Josh, you’re probably right :D, but maybe… both?!
The good news is that the next generation of juniors won’t be allowed to work on real projects thanks to AI and stupid managers. We’re saved!
Again this “not allowed to” non-sense. You’re not required to do things you don’t want to do. If you have a choice between getting a job where you would have to do something you don’t want to do or doing something else or working somewhere else, I would highly recommend taking the job that doesn’t require you to do things you don’t want to do. There is more than one job out there and there is more than one profession. You could even just go live in the woods and eat vegetables all day. Not allowed to? JFC.
Now imagine how useless you are, whatever generation you’re from. Remember, you’re a worthless useless group of individuals who don’t think for themselves, brains rotted out by TV or rock music or whatever.
He won’t have a problem working another 25 years in his role.
Unless he’s hit by a bus, or someone pulls a Luigi on him.
Ah yes, because Linus is topologically identical to a sociopathic insurance CEO for some reason.
Isn’t the point someone would add that extra hole?
Are you suggesting you might try to hurt Linus? He’s done nothing wrong and hurts no one aside from harsh words from time to time. If not you, who on earth are you imagining?
Are we sure that breaking people’s hearts or their self-esteem through mean words don’t hurt them?
Let’s imagine we’ve met our beloved movie star or musician that we adore so much in person
and then he/she says the meanest things, in most vulgar language to us or even laughs at us?
Wouldn’t we be mentally shattered, feel wounded, ashamed or be so crushed and worthless that we consider ending our life in that very moment? Even just for a second?
Because that’s what this guy is. He’s not just a random author of piece of little known software.
He’s a star, a VIP. The leader, the alpha wolf. The Linux godfather himself.
And the mailing lists are public. Everyone can see the victims being dehumanized. Forever. The internet never forgets.
Sounds about right. Dennis Ritchie retired at 66. I thought he’d be an example of working into old age..but since I’m 64 now it doesn’t seem that way.
What about ending Linux development for good and let it rest innpeace and dignity?
I really mean it! If Linux nolonger can’t be maintained the way it used to he, then why not clean up the code one more time and freeze it at at known good condition?
Like a final release, basically. Something like an LTS release?
What speaks against that? A Linux 2.0 using a new name could then be the successor to it.
Full of Rust, A.I. and post-Unix design elements. A fresh start, basically.
No offense, I do admit that I often mentioning harsh criticism on Linux and that I’m not its biggest fan.
But even I have empathy and would’nt want to see it turning into AI slop (thinking of Win 11).
Because Linux is more than just code. It’s a way of life.
Over and beyond, it is sort of a religion to many fine (if kinda misguided) souls that believed in its principles
and invested years (their best years) of their life to improve it.
Those people and their legacy shouldn’t be made superflous through AI.
They deserve to be remembered and be honoured. IMHO.
From Dune (the book, not movie!) “Once men turned their thinking over to machines in the hope that this would set them free. But that only permitted other men with machines to enslave them.”
..written 61 years ago, and here we are.
For now, there’s still NetBSD.
Herbet (senior) was a genius. So far the only visual adaption I’d want to see … is Jodorowsky’s. Just for the sheer weirdness of it.
Yes, definitely Herbert senior. Brian should have left Dad’s legacy alone. I tried a couple of his books long ago. Decent stories. None of his dad’s depth.
“For now, there’s still NetBSD.”
\o/ Yessssss!
I can’t explain why, but NetBSD is probably the only *BSD that didn’t felt too alien to me. I never had the patience to try to understand the whole idea and structure of it (which is only my fault), but my main impression from it for the short time using it (a long time ago I had to setup an old pc as SOHO router), it seemed like a very well organized system (I still remember looking at how simple was to config a NAT, compared to… what was it, ipchains/iptables back then?).
Fun story – when I started learning and using Linux back in ~1995, I started my journey with Slackware (ehh, my first love!) and I’ve heard people saying that it was one of the closest-to-BSD Linux distros.
Don’t be so dramatic. Fork it, roll back the commits. It’s open source.
Meanwhile, there must be dinosaurs living on some island somewhere, because life uhhhh finds a way.
Did you know that fiction isn’t the best source of fact?
Fairytales and ancient mythology do often contain a grain of truth.
They’re both a warning and a reminder for whole generations.
A collective memory of principles, wisdoms and hopes and dreams..
The source of code matters far less than if it is formally verified and maximally energy efficient.
Ironiccally, the use of AI/LLM wastes a lot of energy and water. Among other resources.
It would be interesting to calculate how much energy is saved.:
Linux energy efficiency vs the energy wasted by using AI/LLMs.
IMHO, it reminds me of the old statement that production of solar cells cost more energy than what the solar cells will produce during their entire lifetime.
Thus, as some people say, their only sensible use case is in space/orbit where the sun is an efficient and reliable power source.
No atmosphere, no weather, full radiation spectrum available to the solar cells.
They’re ideal for inner solar system, thus. Satellites, space stations, probes..
Time to reacquaint myself with NetBSD…
Why are we allowing thr GPL Linux Kernel to be infested by not GPL compliant code slop?
Bill, because It’s not up to you and me to decide. Even Linus gets overrun by the hordes of zombies from time to time (rust?).
He’s tired, boss
YOU are allowing it personally because you’re not forking the kernel and doing things on your own. That’s the only way it becomes your personal project. That’s why WE are allowing it too. Well, that’s not why I allow it, but that’s why people who don’t like it are allowing it. It’s open source. If you don’t like the way things are done, fork it and make it better/different/your way. That’s what I did with cpython. Now I don’t have python dependency hell. Simple.
The “Don’t like it, fork it!” argument is a sh*tty one if you ask me. Projects can expand into these beasts that would require expertise and potentially help from others to realistically stand a chance at making a fork. Most simply don’t have that and thus have little choice but either walking away empty-handed or acquiescing.
At that point it stops being an actual option and instead the sentence becomes more an excuse to beat-down on critics be they user or developer for daring to have a differing opinion.
It is one of the more common problems within Open-source. This toxic mandatory-positivity.
No, it’s not a shitty one. You don’t get to dictate what others do.
Hi! That’s true, I suppose.
But isn’t the threat to fork a project also a way of imposing one’s will? 🤔
A kind of emotional blackmail, a form of psychological manipulation? 🙁
I mean, a genuine reason to fork wouldn’t come with a prior announcement and an attached condition, would it?
About that “no one tells me anything!!” way of thinking..
It’s probably not relevant but personally I have a little aversion against this US (?) mindset/mentallity.
Because I believe that in a free society, you don’t simply silence others.
It’s about give and take, about respecting each others.
So sometimes you have to listen to things you don’t agree with.
You may feel being lectured by someone and that bothers you.
That can be uncomfortable, yes—but “listening” and
the willingness to understand the other person do form the foundation of a free, democratic society.
Without that, peaceful co-existence is hardly possible, I’m afraid.
If you can no longer stand it, or if you’ve heard enough, you are free to leave in peace.
Or, better yet, you can speak up and offer constructive counter-criticism. :)
PS: I’m thinking of respectful communication, not that “free speech” thing that involves having the right of insulting others from the very heart.
I mean freedom of opinion/freedom of expression.
Like expressing a point of view in a forum or in a parliament.
This also my consideration, because it undermines the power of team work.
I’m all in for freedom, rights and support of the indiviual but
sometimes individuals need to co-operate and being able to agree on a compromise.
Trouble-solving is part of character development and growing up.
Forking a project due to disagreement/conflict is like leaving a friend in need.
It sometimes can’t be avoided, sadly, but it shouldn’t be the first option.
Or as The Three Muskeeter used to say: “All for one and one for all!” 🙂
https://en.wikipedia.org/wiki/Unus_pro_omnibus,_omnes_pro_uno
https://www.youtube.com/watch?v=1CrqkZk64-w
Because anyone who still thinks it’s slop is mentally deficient.
Because he’s american now and american judges decided that a machine regurgitating another authors work is not the same as a human copycat but also not the same as just literally copying it.
I generally don’t really concern myself with the Linux foundation, as Torvalds has always been a bit of a blunt dicktator of it. Which anyone used to open-source community will know: Is not remotely a rarity. If he declares no stigma of any kind is allowed, then his will: be done.
I just hope people will keep in mind that the Linux foundation has enough resources to actually deal with it, including access being given to cutting-edge models for agents that can effectively filter out immense quantities of c***. It is an extraordinary exception.
Most projects still have to deal with the aftermath of opening Pandora’s box manually and may wish to ward off AI for various reasons.
Note that the Meta-funded LLM only reviews networking-related patches. The overal LLM review is the Google-funded Sashiko project. The Meta money funds a cross-review sashiko instance that’s network only.
Unleash the plagiarism machine onto the Kernel without tagging what parts were written by the plagiarism machine. Brilliant.
Yea, you never learned what plagiarism is. Great.
Good. It’s not 2022 anymore. These models with a right workflow do amazing things now. They can iterate on code and pass review and get code merged better than humans when done right. They can spot exploits and security issues faster and better than humans can. They’re simply becoming an integral part of software development.
And for everyone saying this is some death of society or something. Nothing happened to the chess world when computers dominated chess, except perhaps improved it. In fact, humans started playing at their best AFTER computers took over.
And for those same people, your syntax highlighting, your favorite IDE, etc. were all heralded as the death of competent human programmers, and this is just another chicken little topic like those were.
yeah that’s what i was thinking. personally, i like to code. i don’t want to use AI even if it was very good. and i have cleaned up after some volume of slop from AI, and also slop from human coders.
but the thing driving anyone who is working on practical concerns instead of my own personal preference is exactly waht you say: it’s not 2022 anymore. the capabilities and trade offs are changing so quickly!
I tend to think that’s in big parts because of the huge set of human training data LLMs have (had) access to.
Code snippets on programming forums, repositories, SDKs etc.
What happens if human made content is depleted because humans use LLMs now and nolonger produce training data?
Garbage in, garbage out? Or does AI develop an actual understanding of what it talks about?
But if it does, which might solve this specific issue, what happens next?
How long will it take to decission making and self-awareness? And would we notice?
Personally, I don’t know for sure. But time will tell, I’m afraid.
this is why I think “AI” will ultimately fail. Everything becomes a product of its own invention, it ends up sniffing its own farts. One only needs to look at the amount of generated webs pages on a given topic. They just regurgitate the same information, irrespective of correctness
Hi! What I think AI/LLM might be good for is simulation of historical characters.
It would allow students to ask them questions and the simulations would mimic their speech and their personality.
The public and documented side of personality, at least.
We know little about their private life, after all.
But the AI/LLM could access their biography and their works, at least.
Private letters might fill a small gap, also.
The whole thing would be some sort of advanced role play, like the holograms on ST:TNG, basically. :)
The Linux guys then could have their own Linux/Open Source idols on the desktop!
So they can ask them questions and can enjoy to get offended, as if it were the originals. ;)
What’s also interesting is giving AI/LLM a limited data set of a given epoch.
Such as news magazines and articles of 19th century only.
That way, we could ask an AI/LLM to solve certain problems from the point of view of an bygone era. That would be fascinating!
Maybe we could have an AI/LLM that is stuck in the 1930s, 1950s or 1980s.
Its responding to current topics would be amusing, at very least.
Especially if the language and idioms of a certain era is faithfully being represented.
Afaik some danish research time tried something similar. They aftertrained an LLM to use the 1850ies names of 50 or so birds, and afterwards tried to llok at what “opinions” the LLM had. It was more or less what the training data, i.e. the consumed media from that time, described the world to be.
Shill
As usual, Linus is right. Who cares who or what made the code. Have the maintainers keep an eye on it or restrict the patch set size limit or something to keep it manageble.
So has anyone forked it yet? Surely some dev is as curmudgeonly as me
I can understand dropping the AI tag if maintainers felt it wasn’t giving them much useful information. A patch isn’t automatically good or bad because an LLM helped write it, and someone still has to take responsibility for the code. But I wonder if removing the tag also makes it harder to study whether AI-assisted contributions are creating different kinds of problems over time. The Sashiko approach is interesting, especially using multiple models to check each patch, although several models agreeing doesn’t necessarily mean they’re correct. I’d be interested to see actual numbers on bugs caught, false alarms, and how much time it saves human reviewers. Without that, it’s hard to tell whether this is improving code quality or just speeding up the review process.