[a / b / c / d / e / f / g / gif / h / hr / k / m / o / p / s / t / u / v / vg / vm / vmg / vr / vrpg / vst / w / wg] [i / ic] [r9k / s4s / vip] [cm / hm / lgbt / y] [3 / aco / adv / an / bant / biz / cgl / ck / co / diy / fa / fit / gd / hc / his / int / jp / lit / mlp / mu / n / news / out / po / pol / pw / qst / sci / soc / sp / tg / toy / trv / tv / vp / vt / wsg / wsr / x / xs] [Settings] [Search] [Mobile] [Home]
Board
Settings Mobile Home
/g/ - Technology


Thread archived.
You cannot reply anymore.


[Advertise on 4chan]


File: 5345626.png (567 KB, 1080x1350)
567 KB PNG
Linus has chosen to spit in the face of wise men who came before him. The 'Trusting Trust' problem Thompson demonstrated in his 1984 Turing award lecture proved that a compiler can be modified to insert backdoors into everything it compiles, including future copies of the compiler itself, completely undetectable in the source code. This problem has never been solved. Wheeler's method for Diverse Double Compiling only solves it in theory and relies on the assumption of an already trusted 2nd compiler (Paul Jakma's 2010 paper critiquing DDC goes into more detail). Self-reproducing builds don't even come close because the core issue of the build tools isn't addressed at all. Bootstrapping also falls short because it relies on an existing binary (seed) to advance to the next version of the compiler, the chain of compromise has not been eliminated, the fundamental problem is never confronted in any meaningful way.

Whether you're using Claude Code from the terminal, an IDE with Cursor's models embedded within it or GitHub Copilot; the AI models you're using are upstream from the compiler, with greater privilege. No matter how much control you think you have, the tool you're using has become the developer, the reviewer & the build system. The difference between Ken Thompson's infected compiler and a compromised AI coding assistant, is that the coding tool doesn't need a malicious self-producing binary mechanism. It has access to the .env file, can make changes to a CI/CD pipeline and alter test fixtures. These AI tools have further obfuscated the chain of trust to such an extreme degree, any method of auditing is futile.

Top it off with wilful ignorance towards distillation; training a model on the outputs of another. Private companies with proprietary AI models simultaneously poisoning their data in an effort to stop other AI models from scraping data while depending on AI tooling that trains on the same ecosystem.
>>
Deranged old man.
>>
>>109293428
Good thread anon, too bad that it will be spammed with snail cat memes
>>
You will get insulted for pointing these flaws out
>>
luddites lost lmao
>>
>>109293428
>tl;dr snailcat babble
your trans
>>
>>109293428
>OH MY GOD THIS INTERNET PERSONALITY HAS A NUANCED TAKE ON CURRENT THING!!! THIS IS AN OUTRAGE!!
>>
File: 6098429.jpg (31 KB, 399x501)
31 KB JPG
>>109293556
*you're
glowies stay mad
>>
File: IMG_3066.png (195 KB, 509x694)
195 KB PNG
>>109293428
LinusGODS won and codetrans lost
>>
>>109293514
>>109293556
>>109293573
>>109293630

Ask your favorite AI this question "is it technically possible for an ai coding assistant to insert invisible strings of text into the code that aren't visible when reviewing source code"
It will give you the correct answer and then try to comfort you with suggestions that all amount to the inevitable chain of trust.
>>
>>109293762
Cherry-picking is not real life lil snailcat
>>
>>109293762
>insert invisible strings of text into the code
how is that possible? can you demonstrate it?
>>
>>109293841
Basically, fuck Unicode.
https://arstechnica.com/security/2026/03/supply-chain-attack-using-invisible-code-hits-github-and-other-repositories/
>>
>>109293573
I don't think that's a nuanced take when you're talking about the OS that powers basically fucking everything important on earth except for like the 20 very specific things RTOS is still used.
If linux ever gets compromised it's bringing society down with it.
>>
>>109293855
oh okay, so there needs to already be code in the project that can convert that unicode back into executable code. i thought you meant that the compiler or interpreter was directly executing the invisible code
>>
>>109293892
>i thought you meant that the compiler or interpreter was directly executing the invisible code
A compiler can read strings of text not visible to the human eye as executable code. It's 100% possible for an AI coding assistant to write invisible code, comment out lines of your code that are ran as checks, etc. without you knowing. Go ahead and analyze the binary, what software did you use to do that ? Did you compile that from source ? What about the compiler you used ?
>>
File: ry457.png (92 KB, 1997x716)
92 KB PNG
>>109293923
>A compiler can read strings of text not visible to the human eye as executable code
like i said, it doesn't directly execute the unicode. it needs some obvious decoder code to work
>>
>>109293945
Don't compilers and interpreters decode unicode in order to compile or interpret your code? I'm pretty sure decoding unicode characters is the most necessary part of both of those things or they don't work.
>>
>>109293762
didnt read lol
your trans btw
>>
>>109293945
What's stopping the compiler or command line interpreter from having this feature ?
>>
>>109293979
>Don't compilers and interpreters decode unicode in order to compile or interpret your code?
they don't shift the unicode which you need to do if you want to turn invisible unicode into keywords that the compiler understands
>>109294033
nothing, but that would require you to go install the malicious version of the compiler as well
>>
>>109294047
>that would require you to go install the malicious version of the compiler as well

What if all the major compilers for every language and those maintaining them have been compromised ? How difficult do you think this would be for a nation state with a well funded intelligence agency.
>>
>>109294074
If it were that bad, then it would never have been such a huge deal that your screen could be remotely reconstructed by intercepting rf leakage.
>>
>>109294074
your computer is already compromised by the glowniggers. they have no reason to try hiding the same exact exploit into every single language and somehow make the every team of maintainers complicit
>>
>>109294106
dont waste your time with snailcats
just call em trans, thats all they are
>>
>>109293923
I'll just have my AI check the code for any malicious invisible lines or strange functions that could interpret things as code and warn me or just fix it.
>>
>>109293428
live-bootstrap is getting closer to solving it, but it's still difficult
Guix still hasnt gotten rid of Guile or Linux-libre from their bootstrap just yet
It'll cone in due time
>>
>>109294314
exactly
snailcats dont get theyre obsolete now lool
>>
>>109294106
>they have no reason to try hiding the same exact exploit into every single language and somehow make the every team of maintainers complicit
They don't have to though. AI autonomously builds entire software tools — including compilers — creating a literal realization of Thompson’s attack. In Feb Anthropic demonstrated that 16 parallel Claude agents, running autonomously over nearly 2,000 sessions, built a 100,000 line C compiler in Rust that can compile the Linux kernel on x86, ARM, and RISC-V. No human intervention.

It's less about spying on autists and more about having an edge over nation states and control over the internet. China has their own internet for a reason.
>>
>>109294314
How do you audit the AI you're using to know it's not cucking you
>>
>>109294383
that's cool. thankfully, no one uses it because it is worse than the compilers already installed on peoples systems
>>
>>109294410
How can anyone be certain the compilers already installed on peoples systems aren't compromised ?
>>
>>109294420
you can ask that question about all software
>>
>>109293428
> the AI models you're using are upstream from the compiler, with greater privilege
bruh they just (assist with) writing source code

any of the developers could already have compromised tools somewhere. perhaps their editor or their filesystem driver, who knows. you're really not making any point here.
>>
>>109294420
your trans.
>>
>>109294438
OK so we agree, you and I are brothers bound by cursed knowledge. Whether you like it or not.
>>
>>109294452
how do you audit proprietary AI models tho lol. How can you audit the data it was trained on ?
>>
>>109294570
> How can you audit the data it was trained on ?
Same question as can you audit EVERYTHING a human might have been trained on from the moment of birth? Do you even know this in detail like you at least might know for an AI model you trained?

But also how do you audit the hardware and software humans are working with, all of them potentially malicious?

The answer to all: You don't. You just audit the source code.
>>
>>109294570
How do you audit everything? You can't it's that's simple. Like it or not Ai is a tech that is here to stay. I personally think it's overrated and pretty retard. But this isn't going back in the box.
>>
smart, linus
>>
File: 1784060612919.png (123 KB, 1100x458)
123 KB PNG
It's okay when jeets vibecode for troonix
>>
>>109294689
Vibegods will inherit the earth, tranny
>>
I wonder if he'd keep this opinion if he ever actually looked at the state of the fucking kernel in LTS
>>
>>109293428
with competition between Great Powers, we might actually get an escape hatch from this problem. As LLMs become better at reverse engineering, it might become feasible to actually verify compiler output, keeping the actors in the space honest.
>>
>>109293428
this is why we can't trust open source.
>>
>>109293428
You're a retard.
Shut the fuck up.
>>
File: rms was right.png (934 KB, 869x763)
934 KB PNG
>>109293428
Linus has always been a corpo american big tech sellout.
>>
>>109296031
What was he right about?
>>
>>109296040
>Reasons not to use ChatGPT
>ChatGPT is not "intelligence", so please don't call it "AI".

>I define "intelligence" as being capable of knowing or understanding, at least within some domain. ChatGPT cannot know or understand anything, so it is not intelligence. It does not know what its output means. It has no idea that words can mean anything.

>I call it a "bullshit generator" because it generates output "with indifference to the truth".

>The same applies to many other "generative systems", for the same reasons

>The widespread public error of attributing intelligence to those systems leads millions of people to a misplaced trust for them. Please join me in spreading the word that people should not trust systems that mindlessly play with words to be correct in what those words mean.

>Another reason to reject ChatGPT in particular is that users cannot get a copy of it. It is unreleased software -- users cannot get even an executable to run, let alone the source code. The only way to use it is by talking to a server which keeps users at arm's length.

>Doing your own computing via software running on someone else's server inherently trashes your computing freedom.
>>
>>109296040
Tivoization. Basically the reality of modern smartphones with locked boot loaders that only run manufacturer signed images bypassing the open source nature of the OS. GPLv3 fixed this but Linus declaired it would be too hard to re-license the kernel under it.
>>
>>109294074
AI doesn't have anything to do with that scenario though
>>
>>109296051
>>109296061
>schizo rambling with no point to it
Again: what was he right about?
>>
>>109294334
Guix really might be the endgame of the GNU project if they can get GNU Hurd working seriously. Crazy how relevant they might become in such a short space of time.
>>
>>109296040
RMS was right about absolutely fucking everything.
>>
>>109293428
>some not-a-literal-who says something not anti-AI
>immediately on front pages everywhere
Very organic
>>
>>109296261
>not a single example
So, he wasn't right about anything specific and you're just in his cult.
Got it.
>>
File: 1764266115139568.jpg (62 KB, 469x521)
62 KB JPG
>>109293428
I don't care what this ESL "thinks".
>>
Its so hilarious seeing reddit and /g/tards larping as software engineers
What useful shit has they created again? Tox? Lmao.
>>
First Carmack and now Linux. Its sooo fucking over for snailcats. Their strongest soldier is some guy trying to re-invent C(an ex ok cupid engineer btw, the same company that leaked critical information of 1000s of users way before llms) and had a melty over it when the only useful product written in his language was rewritten in Rust. While buffcats got actual industry innovators who created entire new fields of software.
>>
>>109296253
>Crazy how relevant they might become in such a short space of time.
They will forever be irrelevant. The trannies that this will attract are not going to make shit.
>>
>>109294515
My trans?
>>
>>109293428
if he said any different he'd be sent for another reeducation round, stop worshipping ppl
>>
The collective hivemind on reddit and 4chan suddenly deciding they are better informed than some guy who did more than their combined worthless lives ever did after they expressed an opinion they didn't like will never stop being funny to me.
>>
>>109293428
kek all the tr00ns who were chroming his cock are already turning against him, even if these tr00ns 100% use AI in secret.
>>
>>109298874
stallman won, stalling is the way
>>
>>109298934
I mean if you are happy with how Stallman does computing, its not that bad.
https://stallman.org/stallman-computing.html
I mean, its not even bad, using 20 year old hardware to mostly only read text based web articles. I would say its the far more healthy way of using the internet.
Sadly, thats not how even the trannies who are vehemently anti big tech would want the internet to look like.
>>
Quality OP.
>>
>>109293428
>Bootstrapping also falls short because it relies on an existing binary (seed)
Wouldn't a small enough seed that anyone can read and confirm like in Guix case solve this?
>>
>>109293428
The smartest people are Pro-AI like linus and Terrence tao.
Luddites are made of midwits who fear the machine kek
>>
>>109293428
Back to mastodon
>>
>>109296253
>Crazy how relevant they might become in such a short space of time.
No they won't. The reason Linux is as big as it is today is because its largest contributors are mostly corporates with actual incentive to maintain the project(mainly because their services relies on the project and also because as a counter to Windows).
The image you have it in mind of it being an anti corporate tool maintained by a few enthusiastic underground hackers haven't been true since the 00s.
https://www.reddit.com/r/linuxquestions/comments/za564c/is_it_true_that_huawei_and_google_are_among_the/
>>
>>109301551
It would. The seeds are miniscule and the hex code can be read by a human
>>
>>109293428
kys retard
>>
>>109293841
Just save it to a text file that lets you convert it to ISO, you'll see huge stretches of mysterious blocks.
>>
>>109293858
linux was already totally compromised, all versions, for multiple years. did anything happen? did you nootice? copyfail, just weeks ago.
rhel glowies didnt even patch it for week, makes you think
>>
>>109294097
Maybe they made that a huge deal to distract from this.



[Advertise on 4chan]

Delete Post: [File Only] Style:
[Disable Mobile View / Use Desktop Site]

[Enable Mobile View / Use Mobile Site]

All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.