trollbearpig

@trollbearpig@lemmy.world

• ⁨Comment⁩ on ⁨Secure Boot is completely broken on 200+ models from 5 big device makers⁩ ⁨⁨3⁩ ⁨months⁩ ago⁩:

  Probably to late, but just to complement what others have said. The UEFI is responsible for loading the boot software thst runs when the computer is turned on. In theory, some malware that wants to make itself persistent and avoid detection could replace/change the boot software to inject itself there.

  Secure boot is sold as a way to prevent this. The way ot works, at high level, is that the UEFI has a set of trusted keys that it uses to verify the boot software it loads. So, on boot, the UEFI check that the boot software it’s loading is signed by one of these keys. If the siganture check fails, it will refuse to load the software since it was clearly tampered with.

  So far so good, so what’s the problem? The problem is, who picks the keys that the UEFI trusts? By default, the trusted keys are going to be the keys of the big tech companies. So you would get the keys from Microsoft, Apple, Google, Steam, Canonical, etc, i.e. of the big companies making OSes. The worry here is that this will lock users into a set of approved OSes and will prevent any new companies from entering the field. Just imagine telling a not very technical user that to install your esoteric distro they need to disable something called secure boot hahaha.

  And then you can start imagining what would happen if companies start abusing this, like Microsoft and/or Apple paying to make sure only their OSes load by default. To be clear, I’m not saying this is happening right now. But the point is that this is a technology with a huge potential for abuse. Some people, myself included, believe that this will result in personal computers moving towards a similar model to the one used in mobile devices and video game consoles where your device, by default, is limited to run only approved software which would be terrible for software freedom.

  Do note that, at least for now, you can disable the feature or add custom keys. So a technical user can bypass these restrictions. But this isbyet another barrier a user has to bypass to get to use their own computer as they want. And even if we as technical users can bypass this, this will result in us being fucked indirectly. The best example of this are the current Attestation APIs in Android (and iOS, but iOS is such a closed environment that it’s just beating a dead horse hahahah). In theory, you can root and even degoogle (some) android devices. But in practice, this will result in several apps (banks in particular, but more apps too) not working because they detect a modified device/OS. So while my device can technically be opened, in practice I have no choice but to continue using Google’s bullshit.

  But at least we are stopping malware from corrupting boot right? Well, yes, assuming correct implementations. But as you can see from the article that’s not s given. But even if it works as advertised, we have to ask ourselves how much does this protect us in practice. For your average Joe, malware that can access user space is already enough to fuck you over. The most common example is ransonware that will just encrypt your personal files wothout needing to mess with the OS or UEFI at all. Similarly a keylogger can do its thing without messing with boot. Etc, etc. For an average user all this secure boot thing is just security theater, it doesn’t stop the real security problems you will encounter in practice. So, IMO it’s just not worth it given the potential for abuse and how useless it is.

  It’s worth mentioning that the equation changes for big companies and governments. In their case, other well funded agents are willing to invest a lot of resources to create much sofisticated malware. Like the malware used to attack the nuclear program platns in Iran. For them, all this may be worth it to lock down their software as much as possible. Bit they are playing and entirely different game than the rest of us.

• ⁨Comment⁩ on ⁨Amazon workers narrowly reject union in historic vote⁩ ⁨⁨3⁩ ⁨months⁩ ago⁩:

  Says while citing “studies” from the fraser institute, a libertarian think tank registered as a charity. Fuck them and fuck you for spreading their propaganda.

• ⁨Comment⁩ on ⁨The AI-focused COPIED Act would make removing digital watermarks illegal⁩ ⁨⁨3⁩ ⁨months⁩ ago⁩:

  I figure since big tech spent quite a bit of money building those datasets and since they were built before the law, they will be able to keep using them as long as they don’t add anything new but I can’t be certain.

  This is a very weird assumption you are making man. The quoted text you sent above pretty much says the opposite. It says everyone who wants to train their models wirh copyrigthed data needs to get permission from the copyright holders. That is great for me period. No one, not a big company nor the open source community, gets to steal the work of people prodycing art, code, etc. I honestly don’t get why you assume all the data scrapped before would be exempt.q again, very weird assumption.

  As for ML algorithms having use, of course they have. Hell, pretty much every company I have worked with has used them for decades. But take a look at the examples you provided. None of them requires you or your company scrapping a bunch of information from randoms on the internet. Specially not copyrighted art, literature, or code. And that’s the point here, you are acting like all of that stops with these laws but that’s ridiculous.

• ⁨Comment⁩ on ⁨The AI-focused COPIED Act would make removing digital watermarks illegal⁩ ⁨⁨3⁩ ⁨months⁩ ago⁩:

  So you are saying that content scraped before the law is fair game to train new models? If so it’s fucking terrible. But again, I doubt this is the case since this would be against the interests of the big copyright holders. And if it’s not the case you are just creating a storm in glass of water since this affects the companies too.

  As a side point, I’m really curious about LLM uses. As a programmer the only useful product I have seen so far is copilot and similar tools. And I ended up disabling the fucking thing because it produces too much garbage hahaha. But I’m the first to admit I haven’t been following this hype cycle hahahaha, so I’m really curious what the big things will be. You clearly know so much, so want to enligten me?

• ⁨Comment⁩ on ⁨The AI-focused COPIED Act would make removing digital watermarks illegal⁩ ⁨⁨3⁩ ⁨months⁩ ago⁩:

  My man, I think you are delisuonal hahahaha. You are giving AI way too much credit to a technology that’s just a glorified autocomoplete. But I guess I get your point, if you think that AI (and LLMs in particular hahahaha) is the way of the future and all that, then this is apocalyptic hahahahaha.

  But you are delisuonal my man. The only practical use so far for these stupid LLMs is autocomplete which works great when it works. And bypassing copyright law by pretending it’s producing novel shit. But that’s a whole other discussion, time will show this is just another bubble like crypto hahahaha. For now, I hope they at least force everyone to stop plagiarising other peoples work with AI.

• ⁨Comment⁩ on ⁨The AI-focused COPIED Act would make removing digital watermarks illegal⁩ ⁨⁨3⁩ ⁨months⁩ ago⁩:

  My man, I think you are mixin a lot of things. Let’s go by parts.

  First, you are right that almost all websites get some copyright rights when you post on their platforms. At best, some license the content as Creative Commons or similar licenses. But that’s not new, that has been this way forever. If people are surprised that they are paying with their data at this point I don’t know what to say hahaha. The change with this law would be that no one, big tech companies or open source, gets to use this content for free to train new models right?

  Which brings me back to my previous question, this law applies to old data too right? You say “new data is not needed” (which is not true for chat LLMs that want to include new data for example), but old data is still needed to use the new methods or to curate the datasets. And most of this old data was acquired by ignoring copyright laws. What I get from this law is that no one, including these companies, gets to keep using this “illegaly” acquired data now right? I mean, I’m pretty sure this is the case since movie studios and similar are the ones pushing for this law, they will not go like “it’s ok you stole all our previous libraries, just don’t steal the new stuff” hahahaha.

  I do get your point that the most likely end result is that movie studios, record labels, social media platforms, etc, will just start selling the rights to train on their data and the only companies who will be able to afford this are the big tech companies. But still, I think this is a net possitive (weird times for me to be on the side of these awful companies hahaha).

  First of all, it means no one, including big tech companies, get to steal content that is not theirs or given to them willingly. I’m particularly interested in open source code, but the same applies to indie art and any other form of art outside of the big companies. When we say that we want to stop the plagiarism it’s not a joke. Tech companies are using LLMs to attack the open source community by stealing the code under the excuse of LLMs being transformative (bullshit of course). Any law that stops this is a possitive to me.

  And second of all, consider the 2 futures we have in front of us. Option one is we get laws like this, forcing AI to comply with copyright law. Which basically means we maintain the current status quo for intellectual property. Not great obviously, but the alrtenative is so much worse. Option two is we allow people to use LLMs to steal all the intellectual property they want, which puts an end to basically any market incentives to produce art by humans. Again, the current copyright system is awful. But why do you guys want a system were we as individuals have to keep complying with copyright but any company can bypass that with an LLM? Or how do you guys think this is going to pan out if we just don’t regulate AI?

• ⁨Comment⁩ on ⁨The AI-focused COPIED Act would make removing digital watermarks illegal⁩ ⁨⁨3⁩ ⁨months⁩ ago⁩:

  Maybe I’m missing something, but I don’t understand what you guys mean by “the river cannot be dammed”. The LLM models need to be retrained all the time to include new data and in general to get them to change their behavior in any way. Wouldn’t this bill apply to all these companies as soon as they retrain their models?

  I mean, I get the point that old models would be exempt from the law since laws can’t be retroactive. But I don’t get how that’s such a big deal. These companies would be stuck with old models if they refuse to train new ones. And as much hype as there is around AI, current models are still shit for the most part.

  Also, can you explain why you guys think this would stop open source models? I have always though that the best solution to stop these fucking plagiarism machines was for the open source community to create an open source training set were people contribute their art/text/whatever. Does this law prevents this? Honestly to me this panic sounds like people without any artistic talent wanted to steal the work of artist and they are now mad they can’t do it.

• ⁨Comment⁩ on ⁨Microsoft bans China-based employees from using Android devices for work, mandates switch to iPhones⁩ ⁨⁨4⁩ ⁨months⁩ ago⁩:

  Maybe read the text? It has nothing to do with prefering iOS. It’s just google refused to company with China’s spyware requests some time ago (a broken clock and all that), so it’s literally impossible to use android for this in China. Apple on the other hand is happy to suck China’s dick hahahaha, so it’s the only option. Man, you Apple fans are really brainwashed, this is a bad look for Apple hahahah.