Skip to main content

in reply to ☆ Yσɠƚԋσʂ ☆

The US government cut off access to Mythos because Anthropic marketing claims it's so powerful that it could be misused. If China has a better system, doesn't that obligate companies that believe the marketing to use the Chinese system to find vulnerabilities in their software before somebody else does?
in reply to i_am_not_a_robot

We live in a post-truth post-logic post-reason era.

What ever makes the line go up more. That's the answer.

in reply to MrSoup

I love how you chuds are still butthurt that your color revolution failed. 🤣
in reply to ☆ Yσɠƚԋσʂ ☆

I know what you are talking about, I'm just confused why did you say it. Purely out of context and not respectful.

While I was showing a screenshot of the inner-reasoning behind a censor, you start talking about me being some type of politically stressed when I don't give a heck about your "colors". Just chill and stop randomly attack people you don't even know who are and what thinks.

This entry was edited (3 weeks ago)
in reply to MrSoup

If you knew what I was talking about then there would be nothing to be confused about. If you don't know what a color revolution is or the history of the US trying to overthrow the government in China, then spend the time to educate yourself instead of clowning around on here. Here's some reading you can start with in fact:

in reply to ☆ Yσɠƚԋσʂ ☆

Please stop. Go touch grass. I don't care about your doctrine and what's inside your brain. What I shared has nothing to do about your provocatory comment. I will no further answer to your disrespectful provocations.

Have a nice day.

This entry was edited (3 weeks ago)
in reply to ☆ Yσɠƚԋσʂ ☆

I'm sorry, you're right. But it's too funny to see you get mad at nothing.
in reply to MrSoup

Ultimately, your opinion doesn't matter. China is winning and there's nothing you can do about it.

@yogthos@lemmy.ml is doing you a favor by trying to prepare you for the new world that China is creating. We just want you to understand what is happening; this is hospice care for the US empire. The empire is dying no matter how hard you try to deny the changes to our material reality. If you don't want to listen, that's fine! You'll live in the new world whether you accept it or not.

in reply to queermunist she/her

Guys, are you all hallucinating or something? Who the hell bring up China or USA and its winning or not something? The US empire? The new world? Are you ok?

I'm talking about their implementetion of censoring. Isn't this a tech community?

You guys really seems out of touch. Seems everything about USA vs China or your vision of it. Get out, do stuff with your friends, meet new people and do fun stuff. Don't get mad at people online about something you entirely made up. I didn't talk about anything you pointed.

This entry was edited (3 weeks ago)
in reply to MrSoup

Why are you talking about "their implementetion of censoring"? It has nothing to do with the OP.

The OP is about China matching Anthropic and you can't engage with that topic because you hate China so much.

in reply to queermunist she/her

Why would I hate China? I don't get why are you putting words up in my mouth.

Why are you talking about "their implementetion of censoring"?


Because it's fun. Don't you test systems or play with software?

This entry was edited (3 weeks ago)
in reply to MrSoup

Guys, are you all hallucinating or something?


No. This is just common lemmy.ml behavior.

in reply to MrSoup

Previously. We’ve covered this dozens of times. I’m surprised that you’re surprised given that your account is over two years old.
in reply to davel

I'm sorry, this link doesn't work:

The server returned this error: Error.
in reply to MrSoup


(Nitter addon enabled: Twitter links via https://nitter.privacytools.io)

This entry was edited (3 weeks ago)
in reply to davel

Those YouTube links don't work anymore and in its patreon are not present (at least without paying).

Though it was an interesting read (of the working links).

in reply to MrSoup

It’s an old post. I know the Reddit link is broken because Reddit banned & censored r/TheDeprogram. Reddit censoring socialists is why Lemmy was created in the first place: en.prolewiki.org/wiki/Lemmy#Fo…
in reply to davel

The proletarian encyclopedia? Lol, I wasn't aware of this wiki.
in reply to ☆ Yσɠƚԋσʂ ☆

I guess that’s why they literally need to censor the topic from the model UI?

The knowledge is in the open weights, though, at least for older GLM releases.

in reply to MrSoup

how is this relevant to the post you rabid orientalist dog? your people are only filled with hate, you're disgusting
in reply to ghost_laptop

Chill out man, you don't even know me and you are talking shit about me. I'm no "orientalist dog" and I don't know who you are referring to with "your people".

I'm simply showing an interesting inner-analysis the LLM do when talking about censored stuff. It is always fun to try LLMs limits an see what have been censored and here there is also an inner reasoning which is cut-off.

Learn respect and how to talk to people before making stupid assumptions.

in reply to MrSoup

Previously:

Are you also going to claim that DeepSeek isn’t censored?

You can download DeepSeek and run it yourself to get uncensored answers.

Large Language Models (LLMs) are not truth machines. They are garbage in, garbage out. The input to English-language models are largely English-language texts from Five Eyes countries, with all the disinformation and bias that that entails. So the DeepSeek company is in a “damned if you do, damned if you don’t” situation. They can either refuse to answer certain questions, in which case Western media will accuse them of censorship; or they can answer them, in which case (a) their model will perpetuate Cold War I & Cold War II falsehoods and (b) Western media will parade those false answers around in a victory lap. They chose the former for the cloud version of their app, and the latter for the local version.

in reply to davel

I'm aware of open weights and yes, it is clearly a server-side block (since the text generation is cut-off). If I'm not wrong it was DeepSeek I tried (long time ago) locally offline and it was willing to talk even about this type of topics.
This entry was edited (3 weeks ago)
in reply to mabeledo

Finding them is a prerequisite to exploiting them, and by far the hardest part. Once you know what the exploit is, abusing it is not difficult.
in reply to ☆ Yσɠƚԋσʂ ☆

I disagree. Finding them has been the easy part. We’ve been using static analysers that could find common vulnerabilities for decades now, and LLMs are pretty good at text tokenisation, searching and matching.

Exploiting them is the hard part. Most researchers would tell you that once they suspect there’s something that can be exploited, there’s a laborious process that comes afterwards, consisting mostly of a lot of trial and error, and progressive discovery.

in reply to mabeledo

You're entitled to your opinion, but finding vulnerabilities goes far beyond simply doing static analysis. LLMs are able to find vulnerabilities that emerge from subtle interactions between different features, where things like keys and security credentials aren't handled properly, and finding these by hand in a large codebase is nearly impossible.

The very process of finding these vulnerabilities gives you a path towards making an exploit. And the LLM can actually do this laborious process largely autonomously as well. It can probe a site for example, look at the results, and iterate on them. It's an incredibly effective tool for both finding exploits and testing them out in the wild.

In fact, you can ask piefed devs about their recent security debacle that an LLM exposed and gave a step by step guide for exploiting.

in reply to ☆ Yσɠƚԋσʂ ☆

I know how finding vulnerabilities works. I was using static analysis as an example of why exploiting them is the hard part, something that you don’t seem to disagree with.
in reply to mabeledo

And I gave you a concrete example of how LLMs both find and exploit these vulnerabilities. It's quite evident that your disagreement stems from not having actually used these tools to find vulnerabilities.
in reply to mabeledo

Yes, quite extensively in fact. That's how I found a massive security hole in piefed that I mentioned earlier in fact.
in reply to ☆ Yσɠƚԋσʂ ☆

That’s always the issue with LLMs. One have to be an expert in the topic so they aren’t fooled by their answers.

And don’t get me wrong, they are legitimately useful. But going back to my premise, finding a vulnerability has never been the benchmark, exploiting it is. And if you are good at your craft, you should be perfectly able to spot such vulnerability because you are trained to avoid it. Exploiting it is a whole different thing, and way more complex.

in reply to mabeledo

What I'm saying here is that the way you actually use LLMs is by having them go through the steps of the exploit. It makes a hypothesis and then it tries it, and then you see the result. There's nothing to be fooled by here because the steps it takes either work or they don't.

The reason LLMs are much better at finding these vulnerabilities is because a human can't keep a large codebase in their head all at once. If you look at a project like Lemmy for example, there's a ton of code in it. You have to be an expert in what that code is doing, how the moving pieces relate to each other, and the domain itself to find the exploit. The LLM can zero in on the problems much easier, and actually take the steps to try the exploit. For example, for the case I mentioned with piefed, the issue was very subtle way the oauth token was being misused. It wasn't localized in one place where auth was done, but manifested in a different part of the codebase that relied on it. Something like that would take a lot of dedicated work to find manually.

in reply to ☆ Yσɠƚԋσʂ ☆

in reply to mabeledo

in reply to ☆ Yσɠƚԋσʂ ☆

Maybe you should stop trying to debate a topic you're very clearly not qualified to have an opinion on.


I worked in security for a decade. I think I am entitled to my own opinion, even if you don’t like or don’t understand it.

Anyway, I have made my point. Cheers.

Edit: lol, straight from your link

Linux kernel boss Linus Torvalds has declared the project’s security mailing list has become “almost entirely unmanageable” due to multiple researchers using AI to find bugs and then filling the list with duplicate reports.


Great job guys!

This entry was edited (3 weeks ago)
in reply to mabeledo

worked in security for a decade. I think I am entitled to my own opinion, even if you don’t like or don’t understand it.


Not only do I understand your opinion, I've also spent a lot of time explaining the problems with your claims here.

Also, not sure what the lol here is. What the part you highlighted supports my point which is it's becoming much easier to find exploits, hence why you see more duplicate reports.

Cheers.

in reply to ☆ Yσɠƚԋσʂ ☆

Depends on the exploit. Sometimes it requires physical access to a port with contacts hidden under conformal coating that damages when removed.
in reply to TrippinMallard

The context here is obviously software exploits given that we're talking about LLM finding them.
in reply to ☆ Yσɠƚԋσʂ ☆

That was not obvious to me. LLMs have been used for finding hardware, firmware, RF, software, and social exploits.

RAM side-channel attacks are a good example of software exploits that are harder to exploit than find the vulnerability.

This entry was edited (3 weeks ago)
in reply to TrippinMallard

Sure, you can do all that as well, but the context is an article about cyber security.
in reply to ☆ Yσɠƚԋσʂ ☆

Cybersecurity includes finding hardware, firmware, RF, software, and social exploits.
in reply to TrippinMallard

Again, I'm not disagreeing that you can use LLMs to audit all these things. All I'm saying is that software is by far the easiest place to apply models and actually try out exploits end to end.
in reply to ☆ Yσɠƚԋσʂ ☆

Your original comment was:

Finding them is a prerequisite to exploiting them, and by far the hardest part. Once you know what the exploit is, abusing it is not difficult.
in reply to TrippinMallard

Yes, and as I've repeatedly clarified here, I was making that statment in the context of software.
in reply to ☆ Yσɠƚԋσʂ ☆

I believe your intent. Reponses before your clarification are justified in operating under a general cybersecurity definition, not one tied only to a narrower scope of just software.

Claiming context is obvious is flawed.
Obvious is a local term.
It may have been obvious to you, it was not obvious from the statement itself.

Your internal context should not be treated as globally shared context.