Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

C2PA Cameras Do Not Survive Contact with Reality

da.vidbuchanan.co.uk

Post reply on HN
Print view View original

101–110 of 153 posts

Re: C2PA Cameras Do Not Survive Contact with Reality

#101

Post by akoboldfrying »

hypfer wrote: post #80
akoboldfrying wrote: post #71

Earlier quoted context omitted.

What makes you so certain that this valuable research showing weaknesses in today's systems will render the C2PA concept useless forever? Yes, software LPEs are a risk -- as they are in every nontrivial computer system. New ones will appear, and old ones will be closed in time, as TFA acknowledges. Re hardware attacks: The (neat!) glitch injection attack the author describes in the linked "lighter" page only raises t…

This is yet another of these absolutely caustic takes that come with a veneer of intellectualism, but actually just ignorantly deconstruct reality. Each of them weaponizing the rules of the platform that (for sensible reasons) demand you engage with the strongest interpretation of the message/argument you see. The asymmetry of effort there is unsustainable, and that's exactly the point. No idea how that could be solv…

There is no content in your response. At all. I honestly don't think I've come across a post this devoid of content on this site before.

All I can perceive in it is a generalised hate towards some broader thing that you feel certain I'm consciously aligned with somehow, and that you think is inherently and obviously evil. Because... I don't use my real name on here? (Is your real name "hypfer"?) Because I don't have bricked hardware to back up my opinions?

If you want to convince me or others that something I wrote is wrong, stop hyperventilating and engage with at least one of the specific claims in my post. For your own sake, I'd also suggest editing or outright deleting your original post.

Re: C2PA Cameras Do Not Survive Contact with Reality

#102

Post by Melatonic »

fedpost wrote: post #58

For the time being, I wish more workflows revolved around RAWs. You can pretty much prove the origins of an image sans cryptography by just possessing the image sensor data that you used to produce the final. It's not foolproof but AI image models can't really generate a convincing bayer raw and it's not something anyone is going to spend significant time training.

Exactly

Re: C2PA Cameras Do Not Survive Contact with Reality

#103

Post by lazide »

ainch wrote: post #100
lazide wrote: post #99

Earlier quoted context omitted.

A huge problem with ML training is the ‘ouroboros issue’ - training ML with input from other MLs breaks things in very deep (but difficult to stop/detect when it’s happening) ways due to the way the internal math works (model collapse). Right now, the Internet training set is becoming more and more contaminated with better and better generative AI images and video. It makes the models more screwed up, and makes it ve…

"Model collapse" is often overstated, as this paper demonstrates: https://arxiv.org/pdf/2404.01413 The original model collapse paper assumes you train networks on 100% synthetic data produced by the previous generation. But if you maintain some portion of real data then the problem is mitigated.

Not when you also including poisoning attacks.

I remember the original paper showing issues with even a couple percent of certain kinds of synthetic data too, not 100%.

Re: C2PA Cameras Do Not Survive Contact with Reality

#105

Post by lbriner »

Gigachad wrote: post #59
mistercow wrote: post #40

Aside from the fact that this was obviously never viable and the entire problem is clearly unsolvable if you sit down and really probe it for fifteen minutes, what I find most frustrating about this is that the false promise of preserving photos as reliable evidence is actively harmful. You will not build a perfect system, or even something near perfect. The best you're going to do is make it so that it's hard to cas…

I don’t think there is a technical solution to this problem but I think there is a legal one. Make it a legal requirement to mark AI generated photos and enforce penalties for posting unmarked AI generations. Social media should also mark the country of origin for each post, with the knowledge that posts from your own country are covered by these laws.

> Enforce penalties

Really, everyone should know by now it is not as simple as that. We already have lots of laws that are not enforced (fully or at all) because there is a cost to enforcing the law - time and money. Some random person uploads an AI image to Facebook and you really expect the police to expend a few thousand $ prosecuting them for what? To Server as a warning?

Some people will be untraceable. Some will claim innocence or mitigation or some disability that excuses them. Even those that might be prosecuted will be fined and won't pay which costs even more to follow up.

We should know that the legal system does not create an obedient society.

Re: C2PA Cameras Do Not Survive Contact with Reality

#106

Post by dTal »

fightfake-ai wrote: post #88
fwipsy wrote: post #57

Earlier quoted context omitted.

I think it's useful even if it can be spoofed. Many people don't even bother to edit visible watermarks out of AI photos/videos. I'm fairly certain this will defeat 99.9% of malicious users, many of whom won't even know it exists until someone points out it's missing. People are concerned that the technology will lend additional credence to the last 0.1%. But anyone who thinks about the technology for 2 minutes will…

I agree with "I'm fairly certain this will defeat 99.9% of malicious users". Also, note that C2PA should have something like Level 3 as well: "The image is mathematically proven to have come from the physical camera sensor." It's somehow difficult to achieve this, but it's possible (although the attacks will always be possible of course).

Difficult and also solves nothing, since you can just point the camera at a screen.

Re: C2PA Cameras Do Not Survive Contact with Reality

#107

Post by hypfer »

akoboldfrying wrote: post #101
hypfer wrote: post #80

Earlier quoted context omitted.

This is yet another of these absolutely caustic takes that come with a veneer of intellectualism, but actually just ignorantly deconstruct reality. Each of them weaponizing the rules of the platform that (for sensible reasons) demand you engage with the strongest interpretation of the message/argument you see. The asymmetry of effort there is unsustainable, and that's exactly the point. No idea how that could be solv…

There is no content in your response. At all. I honestly don't think I've come across a post this devoid of content on this site before. All I can perceive in it is a generalised hate towards some broader thing that you feel certain I'm consciously aligned with somehow, and that you think is inherently and obviously evil. Because... I don't use my real name on here? (Is your real name "hypfer"?) Because I don't have…

It is correct that I have completely side-stepped your frame and not even attempted to engage with it. This is on purpose, because wanting people to engage with that specifically engineered frame is exactly the bad faith mechanism I've described.

What is interesting is what happens when these tactics are called out in a meta comment like I did, because it's actually a common thing that people start flailing like this once you side-step the script that was supposed to be followed.

Someone acting in good faith would not counter with an attack and a double-down like this, but with intent to resolve the situation (+ probably feeling a bit bad about an exchange having failed and being misunderstood).

I also like the "hyperventilating" claim. Easiest way to get out of a situation is to invalidate the source, and easiest way to do that is to claim emotions.

And the fake concern (for what, even?) of course.

> For your own sake, I'd also suggest editing or outright deleting your original post.

___

Anyway. I think the convincing people is actually working quite well here. Though not in the sense you'd think.

I'm just pointing a spotlight at what I believe is violating the spirit of the rules (while staying within the letter of the rules).

What other people make out of that is their decision to make.

Re: C2PA Cameras Do Not Survive Contact with Reality

#108

Post by dTal »

hypfer wrote: post #80
akoboldfrying wrote: post #71

Earlier quoted context omitted.

What makes you so certain that this valuable research showing weaknesses in today's systems will render the C2PA concept useless forever? Yes, software LPEs are a risk -- as they are in every nontrivial computer system. New ones will appear, and old ones will be closed in time, as TFA acknowledges. Re hardware attacks: The (neat!) glitch injection attack the author describes in the linked "lighter" page only raises t…

This is yet another of these absolutely caustic takes that come with a veneer of intellectualism, but actually just ignorantly deconstruct reality. Each of them weaponizing the rules of the platform that (for sensible reasons) demand you engage with the strongest interpretation of the message/argument you see. The asymmetry of effort there is unsustainable, and that's exactly the point. No idea how that could be solv…

That's a very long-ass way of saying "I disagree with this comment, I can't be bothered explaining why (too much work), so instead I think this user should be banned". I extracted nothing useful from your comment, which bears no relation to its parents that I can discern, am irritated to have spent time reading it and composing this reply, and also find your attitude pretty rich when you're also an anon account, and with less karma too. If you want to "call someone out" you've gotta be on damn solid ground and bring receipts. As it is you've just decreased the signal/noise ratio of the thread and contributed to a toxic atmosphere. I would just flag and move on but in this case I think feedback is important, and unlike you I can be bothered to explain what it is that's got me riled up.

Re: C2PA Cameras Do Not Survive Contact with Reality

#109

Post by ainch »

lazide wrote: post #103
ainch wrote: post #100

Earlier quoted context omitted.

"Model collapse" is often overstated, as this paper demonstrates: https://arxiv.org/pdf/2404.01413 The original model collapse paper assumes you train networks on 100% synthetic data produced by the previous generation. But if you maintain some portion of real data then the problem is mitigated.

Not when you also including poisoning attacks. I remember the original paper showing issues with even a couple percent of certain kinds of synthetic data too, not 100%.

What do you mean by poisoning attacks - stuff like Nightshade or Glaze? I was under the impression that those have largely failed to achieve their goals.

Re: C2PA Cameras Do Not Survive Contact with Reality

#110

Post by hypfer »

dTal wrote: post #108
hypfer wrote: post #80

Earlier quoted context omitted.

This is yet another of these absolutely caustic takes that come with a veneer of intellectualism, but actually just ignorantly deconstruct reality. Each of them weaponizing the rules of the platform that (for sensible reasons) demand you engage with the strongest interpretation of the message/argument you see. The asymmetry of effort there is unsustainable, and that's exactly the point. No idea how that could be solv…

That's a very long-ass way of saying "I disagree with this comment, I can't be bothered explaining why (too much work), so instead I think this user should be banned". I extracted nothing useful from your comment, which bears no relation to its parents that I can discern, am irritated to have spent time reading it and composing this reply, and also find your attitude pretty rich when you're also an anon account, and…

This is what I'd call the "enabler" archetype that gets unhappy about friction (rightfully!) but takes some.. uh.. mental shortcuts when it comes to tracing back the source of it.

I'm sorry man. The problem is that this exact described dynamic (see also the other reply by me which you seem to have conveniently missed), is that it affects _all_ internet spaces and makes them utterly miserable.

I agree though that this meta stuff doesn't exactly improve SNR, and I am open for better solutions. But those need to actually solve the problem, instead of just silencing the immune response.

___

Interesting side-thought

> with less karma too

I think this might point at (one of the) root cause(s) of the dysfunction I'm pointing at here.

Internet points ceased to be a metric documenting value quite a while ago. Something something Goodhart

Post reply on HN