[HN Gopher] Watermark Anything
       ___________________________________________________________________
        
       Watermark Anything
        
       Author : zerojames
       Score  : 117 points
       Date   : 2024-11-12 08:08 UTC (1 days ago)
        
 (HTM) web link (github.com)
 (TXT) w3m dump (github.com)
        
       | turbocon wrote:
       | Is this a big deal? I'm a layman here so this seems like a needed
       | product but I have a feeling I'm missing something.
        
         | clueless wrote:
         | this is one of the primary communication methods of oversea
         | agents in CIA, interesting to have it be used more broadly
         | </joke>
        
           | bagels wrote:
           | Do you have a source? I'd be interested in reading more about
           | this.
        
             | edm0nd wrote:
             | of course not, check their username.
        
             | cozmorado wrote:
             | It's a form of Steganography
             | https://en.wikipedia.org/wiki/Steganography
        
         | mintplant wrote:
         | My assumption is that this will be used to watermark images
         | coming out of cloud-based generative AI.
        
           | jsheard wrote:
           | And they'll say it's to combat disinformation, but it'll
           | actually be to help themselves filter AI generated content
           | out of new AI training datasets so their models don't get
           | Habsburg'd.
        
             | muppetman wrote:
             | I wondered why they'd be doing this NOW and this makes
             | perfect sense!!
        
             | Y_Y wrote:
             | > their models don't get Habsburg'd.
             | 
             | You mean develop a magnificent jawline, or continue to
             | influence Austrian politics?
        
           | Animats wrote:
           | Why? Those are not copyrightable.
        
             | xd1936 wrote:
             | Because downstream consumers of the media might want to
             | know if an image has been created or manipulated by AI
             | tools.
        
               | Jerrrrrrry wrote:
               | hardware self-evident modules to essentially sign/color
               | all digital files
               | 
               | now we need more noise
        
             | Moru wrote:
             | They would not want to train their next model on the output
             | of the previous one...
        
         | Jach wrote:
         | Various previous attempts at invisible/imperceptible/mostly
         | imperceptible watermarking have been trivially defeated, this
         | attempt claims to be more robust to various kinds of edits.
         | (From the paper: various geometric edits like rotations or
         | crops, various valuemetric edits like blurs or brightness
         | changes, and various splicing edits like cutting parts of the
         | image into a new one or inpainting.) Invisible watermarking is
         | useful for tracing origins of content. That might be copyright
         | information, or AI service information, or photoshop
         | information, or unique ID information to trace leakers of video
         | game demos / films, or (until the local hardware key is
         | extracted) a form of proof that an image came from a particular
         | camera...
        
           | mananaysiempre wrote:
           | ... Ideal for a repressive government or just a mildly
           | corrupt government agency / corporate body to use to identify
           | defectors, leakers, whistleblowers, or other dissidents.
           | (Digital image sensors effectively already mark their output
           | due to randomness of semiconductor manufacturing, and that
           | has already been used by abovementioned actors for the
           | abovementioned purposes. But that at least is difficult.)
           | Tell me with a straight face that a culture that produced
           | Chat Control or attempted to track forwarding chains of chat
           | messages[1] won't mandate device-unique watermarks kept on
           | file by the communications regulator. And those are the more
           | liberal governments by today's standards.
           | 
           | I'm surprised how eager people are to build this kind of
           | tech. It was quite a scandal (if ultimately a fruitless one)
           | when it came out colour printers marked their output with
           | unique identifiers; and now that generative AI is a thing
           | stuff like TFA is seen as virtuous somehow. Can we maybe not
           | forget about humans?..
           | 
           | [1] I don't remember where I read about the latter or which
           | country it was about--maybe India?
        
             | baltimore wrote:
             | > ... for a repressive government ...
             | 
             | Why shouldn't a virtuous and transparent government (should
             | one materialize somehow, somewhere) be interested in
             | identifying leakers?
        
               | Jerrrrrrry wrote:
               | Both can be true! This is essentially making it easier to
               | do [x] argument, which itself is essentially security
               | through obscurity.
               | 
               | It was always possible to do watermark everything: any
               | nearly-imperceptible bit can be used to encode data that
               | can be used overtly.
               | 
               | Now enabling everyone everywhere to do it and integrate
               | it may have second-order effects that were opposite of
               | one's intention.
               | 
               | It is very convenient thing, for no one to trust what
               | they can see. Unless it was Validated (D) by the Gubmint
               | (R), it is inscrutable and unfalsifiable.
        
               | palata wrote:
               | The parent comment says that it has dangerous use-cases,
               | not that it does not have desirable ones.
        
               | mananaysiempre wrote:
               | That's like asking why a fair and just executive
               | shouldn't be interested in eliminating the overhead of an
               | independent judiciary. Synchronically, it should.
               | Diachronically, that's one of the things that ensures
               | that it _remains_ fair and just. Similarly for
               | transparency and leakers, though we usually call those
               | leakers "sources speaking on condition of anonymity" or
               | some such. (It does mean that the continued transparency
               | of a modern democratic government depends on people's
               | continued perpetration of--for the most part--mildly
               | illegal acts. Make of that what you will.)
        
               | willcipriano wrote:
               | If they are transparent, what is leaking?
        
             | Jach wrote:
             | I stopped myself from making the printer analogy, but of
             | course it's relevant, as is the fact that few seem to care.
             | I personally hope some group strikes back to sanitize
             | images watermarked this way, with no more difficulty than
             | removing exif data.
        
       | wkirby wrote:
       | Link to the paper in the README is broken. I _believe_ this is
       | the correct link to the referenced paper:
       | https://arxiv.org/abs/2411.07231
        
         | Jaxan wrote:
         | There is some nice information in the appendix, like:
         | 
         | "One training with a schedule similar to the one reported in
         | the paper represents [?] 30 GPU-days. We also roughly estimate
         | that the total GPU-days used for running all our experiments to
         | 5000, or [?] 120k GPU-hours. This amounts to total emissions in
         | the order of 20 tons of CO2eq."
         | 
         | I am not in AI at all, so I have no clue how bad this is. But
         | it's nice to have some idea of the costs of such projects is.
        
           | svilen_dobrev wrote:
           | so say i have a site with 3000 images, 2M pixel each. How
           | many GPU-months it would take to mark them? And, what
           | gigabytes i would have to keep for the model?
        
             | hnuser123456 wrote:
             | That amount of compute was used for training. For inference
             | (applying the watermarks), hopefully no more than a few
             | seconds per image.
             | 
             | Llama 3 70B took 6.4M GPU hours to train, emitting 1900
             | tons of CO2 equivalent.
        
               | Jaxan wrote:
               | Thanks! I was not at all aware of the scale of training!
               | To me those are crazy amounts of gpu time and resources.
        
               | pierrefdz wrote:
               | The amounts of gpu time in the paper are for all
               | experiments, not just training the last model that is OSS
               | (which is usually reported). People don't just oneshot
               | the final model.
        
             | GaggiX wrote:
             | The embedder is only 1.1M parameters, so it should run
             | extremely fast.
        
               | pierrefdz wrote:
               | Yes, although the number of parameters is not directly
               | linked with the flops/speed of inference. What's nice
               | about this AE architecture is that most of the compute
               | (message embedding, and merging) is done at low
               | resolution, same idea as behind latent diffusion models
        
           | azinman2 wrote:
           | It's very interesting this is gpu time based because:
           | 
           | 1. Different energy sources produce varyings of co2
           | 
           | 2. This likely does not include co2 to make the GPUs or
           | machines
           | 
           | 3. Humans involved are not added to this at all, and all of
           | the impact they have on the environment
           | 
           | 4. No ability to predict future co2 from using this work.
           | 
           | Also if it really matters, then why do it at all? If we're
           | saying hey this is destroying the environmental and care,
           | then maybe don't do that work?
        
             | TeMPOraL wrote:
             | > _1. Different energy sources produce varyings of co2_
             | 
             | Yes.
             | 
             | > _2. This likely does not include co2 to make the GPUs or
             | machines_
             | 
             | Definitely not, nobody does that.
             | 
             | Wish they did, in general I feel like a lot of beliefs
             | around sustainability and environmentalism are wrong or
             | backwards precisely because embodied energy is discounted;
             | see e.g. stats on western nations getting cleaner, where a
             | large - if not primary - driver of improved stats is just
             | outsourcing manufacturing, so emissions are attributed to
             | someone else.
             | 
             | Anyway, embodied energy isn't particularly useful here.
             | Energy embodied in GPUs and machines amortizes over their
             | lifetimes and should be counted against all the things
             | those GPUs did, do and will do, of which the training in
             | question is just a small part. Not including it isolates
             | the analysis to contributions from the specific task per
             | se, and makes the results applicable to different
             | hardware/scenarios.
             | 
             | > _3. Humans involved are not added to this at all, and all
             | of the impact they have on the environment_
             | 
             | This metric is so ill-defined as to be arbitrary. Even more
             | so with conjunction with 2, as you could plausibly include
             | a million people into it.
             | 
             | > _4. No ability to predict future co2 from using this
             | work._
             | 
             | Total, no. Contribution of compute alone given similar GPU-
             | hours per ton of CO2eq, yes.
        
               | _Algernon_ wrote:
               | >Definitely not, nobody does that.
               | 
               | Except every proper Life-cycle assessment on carbon
               | emissions ever.
        
               | Jerrrrrrry wrote:
               | >proper
               | 
               | doing Scotsman-like lifting when the point was that these
               | things are not considered, or are "externalities"
        
               | albumen wrote:
               | not sure how that invalidates Algernon's point. These
               | things should be considered, and are in a lot of LCAs.
        
               | Jerrrrrrry wrote:
               | >should be considered, and are
               | 
               | Not as much as they should be, was his point. Saying
               | something is not proper is the No True Scotsman fallacy.
        
             | pierrefdz wrote:
             | 1. yes, this is the default co2 eq/ watts from the tool
             | that is cited in the paper, but it's actually very hard to
             | know the source of energy that aliments the cluster, so the
             | numbers are only an order of magnitude rather than "real"
             | numbers 2. 4. I found that
             | https://huggingface.co/blog/sasha/ai-environment-primer
             | gives a good broad overview (not only of the co2 eq, which
             | is limited imo) of AI environmental impact
             | 
             | > Also if it really matters, then why do it at all? If
             | we're saying hey this is destroying the environmental and
             | care, then maybe don't do that work?
             | 
             | Although it may not the best way to quantify it, it gives a
             | good overview of it. I would argue that it matters a lot to
             | quantify and popularize the idea of such sections in any
             | experimental ML papers (and should in my opinion be the
             | default, as it is now for the reproducibility statement and
             | ethical statement). People don't really know what an AI
             | experiment represents. It may seem very abstract since
             | everything happens in the "cloud", but it is pretty much
             | physical: the clusters, the water consumption, the energy.
             | And as someone who works in AI, I believe it's important to
             | know what this represents, which these kinds of sections
             | show clearly. It was the same in the DINOv2 paper or in the
             | Llama paper.
        
           | barbazoo wrote:
           | > This amounts to total emissions in the order of 20 tons of
           | CO2eq.
           | 
           | That's about 33 economy class roundtrip flights from LAX to
           | JFK.
           | 
           | https://www.icao.int/environmental-
           | protection/Carbonoffset/P...
        
             | kube-system wrote:
             | 33 seats on a flight maybe. It's about one passenger
             | aircraft flight, one way.
        
       | Onavo wrote:
       | What if the watermark becomes a latent variable that's indirectly
       | learnt by a subsequent model trained on its generated data? They
       | will have to constantly vary the mark to keep it up to date. Are
       | we going to see Merkle tree watermark database like we see for
       | certificate transparency? YC, here's your new startup idea.
        
         | thanksgiving wrote:
         | I think there should be an input filter that if it sees a
         | watermark refuses to use that input and continues with the next
         | input
        
         | jerf wrote:
         | There's many reasons why people are concerned about AI's
         | training data becoming AI generated. The usual one is that the
         | training will diverge, but this is another good one.
        
         | nickpinkston wrote:
         | I can imagine some kind of public/private key encrypted
         | watermark system to ensure the veracity / provenance of media
         | created via LLMs and their associated user accounts.
        
       | matrixhelix wrote:
       | Now we need a link to the "Unwatermark Anything" repo
        
         | pierrefdz wrote:
         | https://github.com/XuandongZhao/WatermarkAttacker
        
       | rodneyg_ wrote:
       | Does this watermark still work if someone screenshots an image?
        
         | dangoodmanUT wrote:
         | try running the code to find out
        
       | yawnxyz wrote:
       | oh I was kind of hoping this would also watermark text
       | imperceptibly... alas this doesn't do that
        
       | doctorpangloss wrote:
       | I wonder what will come of all the creative technologists out
       | there, trying to raise money to do "Watermarking" or "Human
       | Authenticity Badge," when Meta will just do all the hard parts
       | for free: both the technology of robust watermarking, and
       | building an insurmountable social media network that can adopt it
       | unilaterally.
        
       ___________________________________________________________________
       (page generated 2024-11-13 23:00 UTC)