Hacker News Reader

Comment by Tiberium | Hacker News Reader

It's baffling that this requires auth. OpenAI's own tool (https://openai.com/research/verify/) doesn't require auth.

Even Google themselves previously offered a no-auth way (just very niche): if you uploaded an image to Google Image search, and went to "About this image", it would show if the image was Google AI-generated. Now it doesn't.

View on HN

These are "spy"marks, not watermarks:

https://brand.io/article/spymarks/

We need to inform everyone that this technology can encode database identifiers and enough entropy to uniquely identify you as an author (or downloader).

This extends to other forms of media as well.

They are not spymarks if they don't encode any personal IDs and are merely used to indicate that an image was AI-generated. I don't think Google or OpenAI use SynthID to include personal data.

And there lies the rub. You can't verify if they do or don't embed such content. These companies are just like "trust me bro".

Indeed and Google has built its empire through tracking users across the internet so it is really hard to trust that they won't..

It hasn't. Their DoubleCkick/AdSense (embedded website ads) used tracking, but Google built its "empire" through AdWords (keyword ads in Google Search), not AdSense.

Google built it's empire with PageRank which made their search engine popular enough that they could make money on AdWords.

The spying and tracking started pretty quickly though and they've repeatedly demonstrated that they can't be trusted. They've even violated the law on several occasions in efforts to collect data they had no right to.

"If you have something that you don't want anyone to know, maybe you shouldn't be doing it in the first place." - Eric Schmidt CEO of Google

Come on man, they're both advertising companies. Would you resist the opportunity to build that social graph and detect a known user anywhere on the internet?

Of course. Privacy is a feature that increases their product value. Moreover, they could get into trouble with the GDPR if they secretly included tracking information in their images.

So the obvious move is that marketing materials sing how it’s super private, while actual steganography encodes everything to uniquely identify you. If someone discovers something wave it away (and crack down on pesky whistleblowers), eat the fines if the worst happens.

That's not an obvious move, that's an non-obvious, cynical, "let's assume they are evil and don't care about heavy fines" interpretation.

Which is 100% consistent with Google and OAI behavior and nature

Let’s check.

- Google. The canonical “don’t be evil” meme origin.

- Meta. Adjacent “they trust me, dumb fucks” meme origin.

- Apple. Least of all, but still fits: PR pushes double on privacy, but their interests are not their end-users, they just happen to align on some points as they’re not in adtech business.

- Microsoft. The classic case for getting slaps on the wrist, Gates said one thing he’d do differently is sending lobbyists earlier, or something to that effect, didn’t he?

- Amazon, Netflix, Uber, Airbnb and many more are still all variants of the same story. It’s not a lie that everyone nowadays expects most megacorps to fuck people over if there’s money in that.

That’s the Order of “Feel for It Again”, with oak leaf clusters already, no? Something’s cursed in this industry, every company that sells software to individual people ends up the same when they become gigantic and endowed by wealth and power, even if they were decent in their early days. They all become those blind heralds of paperclip^W shareholder value(tm) maximizers, and I’m finding myself struggling looking for an exception.

Am I to believe this time it’s going to be any different, to set those - not baseless, I’d say - cynical expectations aside? Any signs there’s a change?

So, I disagree - it’s obvious, given the ad-adjacent end-user-facing megacorp context, OpenAI is right there. Cynical - sadly so, yes. But I’m sure it’s obvious: the idea of such a possible direction comes up pretty quickly and reliably so.

> Moreover, they could get into trouble with the GDPR if they secretly included tracking information in their images.

Do you think that matters to Google? They've been fined repeatedly for violating GDPR already. It's like multiple violations and fines every single year going back to 2019. They don't care about breaking the law if it gets them what they want. They've been found guilty of breaking the law in multiple countries, including the US. They've easily been able to afford every fine and still come out massively profitable. Google is above the law and they know it.

>I don't think Google or OpenAI use SynthID to include personal data.

Why? I assume by default that it's uniquely identifiable, because it just makes practical sense for OAI and Google (tracking misuse at the very least, but also data), it's trivial to implement and trivial to hide or plausibly deny.

I don't think it's trivial implement. Encoding more bits is harder or more brittle, and not trivial to hide. One could just ask for a completely white image. Any image distortions must then be due to SynthID. If the distortions differ for different accounts, that's strong evidence for varying watermarks.

Making the unique ID a fixed part of the seed is already enough. The spymark doesn't need to be fixed or 100% reliable.

Cool post, but I'm wondering what would a realistic way to avoid this?

It's not really possible to ban steganography, is it?

> That future lies halfway between now and 1984. So let’s stay off that timeline, shall we?

> Call a spymark what it is. A spy tool used to spy on you and everyone you interact with.

You can't stop people from using steganography, but when they offer it as a feature to track people you can at least reject their preferred marketing term and call it by a name that reflects what it really is.

I'm confused why no one came up with this terminology for stenographic watermarks before. Now that there's an actual prosocial use for them (helping encourage clear labelling of AI-generated content), suddenly people are coining negatively-loaded terms for it, even though the technology has been in use for nefarious purposes for ages.

It's presumably so they can rate-limit people who are trying to reverse-engineer or otherwise strip it (e.g. iteratively tweaking until it stops getting detected)

I wonder how relevant this really is.

How hard can it be to download a set of real images and generated ones in order to train a model to detect and strip the watermark with minimal perceptual difference?

> iteratively tweaking until it stops getting detected

I don't think that dodgy folks have any issue with registering thousands of IDs. I know that I used to regularly get approached by these Chinese companies that were selling good reviews of my free apps. They did this by having thousands of legit AppleID accounts that would download and rate the app.

I haven't been approached by one of these in a long time. I guess Apple figured out how to put the kibosh on them.

It's worth nothing that this IS Google.

Given that it's an ad company, they want something to connect what you're doing to the rest of their data, and they couldn't possibly keep the old image search results there if they want people to use this.

Any access to a watermark detector can be used to remove the watermark. Presumably they want to be able to track who is doing that.

Yeah I've been wanting to run a ton of Wikipedia images through the detector to see if fakes snuck in. Doesn't seem to be a practical way to do this. Even Open AIs tool has a low rate limit

OpenAI's tool is worse though, because it doesn't detect SynthID from non-OpenAI models. I just tried it with various images created by Imagen / Nano Banana.

The EULA you have to accept hints that they are afraid you are going to abuse it to remove synthID - e.g, set up and edit and check loop

That’s the justification, but the EULA incorporates Google’s regular consumer terms; which means what you upload can also be used for advertising and targeting; and basically any purpose whatsoever by Google.

Ah, yes, the mighty EULA. Bad actors always kowtow to these.

I've never seen it mentioned anywhere so I will just here complain that Google also removed the ability to paste images into Google Image search some years ago for no apparent reason. It worked great and I used it all the time.

this + image translate with copy+paste is the only reason I ever use yandex of all places...

I might be missing something but Google Image Search still works for me.

I wonder if it’s a deliberate anti-user decision to get more URLs and less pasted images.

URLs expand Googlebot’s indexes; pasted images don’t.

If they wanted more urls they would allow mobile users to search by url instead of only allowing them to search by image.

Click the little camera icon in the search box and it pops a modal that says Search any image with Google Lens.

The textbox below it says Paste image link, but you can actually paste an image from your clipboard here, too.

You’re a hero. So glad I complained. Why would they not just keep this working with the main text box focused…

No idea. I used it a lot too, not sure if pasting here worked from the beginning of them introducing this Google Lens thing or not.

In similar fashion in the Android Play Store they moved the search off from the top bar. For some time they educated users that "oo the search is no longer here, you need to click there -> to get to the page with search!". That has stopped (for me?), but they still haven't found any other use for same place in the first view.

I suppose it was too convenient.

And if it doesn't work on the Google home page, just go to images.google.com and do the same, it always works there for some reason.

It's deliberate so that you can't find a way to bypass it.

Reply on news.ycombinator.com