#glaze uchicago

[bethanyberg]

GLAZE 2.0 IS OUT TODAY FOR DOWNLOAD! Go protect your art!

NEW: Glaze 2.0 (Apr 14) We are excited to announce the release of version 2.0 of Glaze. This new version significantly improved Glaze robustness against the newest AI models, and requires less time to glaze images. Most notably:

Significantly improved robustness against Stable Diffusion 1, 2, SDXL, especially for smooth surface art (e.g. anime, cartoon).

Less noticeable modification to images, addressed non-convergent patch artifacts

Improved computational efficiency on most platforms (~ 50% speed up)

Built in Mac GPU support -- Over 5X speed up on Mac M-CPUs.

[vanyamired]

Good news, fellow artists! Nightshade has finally been released by the UChicago team! If you aren't aware of what Nightshade is, it's a tool that helps poison AI datasets so that the model "sees" something different from what an image actually depicts. It's the same team that released Glaze, which helps protect art against style mimicry (aka those finetuned models that try to rip off a specific artist). As they show in their paper, even a hundred poisoned concepts make a huge difference.

(Reminder that glazing your art is more important than nighshading it, as they mention in their tweets above, so when you're uploading your art, try to glaze it at the very least.)

[uchicagomagazine]

Meet our Spring/24 UChicagoan, computer science professor and developer of Glaze and Nightshade Ben Zhao! Read the full Q&A here: https://mag.uchicago.edu/science-medicine/ben-zhao

Illustration by Daniel Hertzberg

[made-music]

Fascinating crew of folks together on this:

If the trajectory is anything like it is/was for visual art, the plea here may very well be too late. Generally, computer-generated “content” isn’t really an issue… THE ISSUE is the data sets used to train the algorithms.

Our stance: If one has not provided express consent for one’s work to be used in data training sets then plagiarism has occurred.

Having a kind of insider knowledge of how these “AI” and algorithms work, I’m confident in that stance.

Here’s hoping for the best… but the world tends to let down those who simply hope….

In the meantime, visual artists can and SHOULD disrupt the development of these algorithms as much as they can.

Useful tools like GLAZE and NIGHTSHADE exist and should be used whenever possible.

Link below.

Brilliant work by the crew at University of Chicago.

https://nightshade.cs.uchicago.edu/whatis.html

[vangoghcoffeeco]

I would love to glaze and nightshade my art but the program is such a pain in the ass, and the process of getting access to the web version is a bigger pain in the ass, to where i don't even want to do it. fuck ai, but i need something i can comfortably use.

[ladyjmontilyet]

is there Glaze or Nightshade for videos or audio yet? thinking of getting back into YouTube but after seeing another creator from my city threatened with being turned into an AI simulacrum if he didn't upload i am... concerned

[sepdet]

How to disable AI scraping your tumblr blogs

TL;DR: Blog Settings > Visibility > Turn off third-party sharing

Mobile:

Desktop:

Here's the Official Tumblr Announcement

UChicago's free tool Nightshade poisons AI image generators with disruptive data

Glaze blocks AI from detecting/mimicking an artist's distinctive style

Fawkes disrupts facial recognition scans

[smellslikebot]

"how do I keep my art from being scraped for AI from now on?"

if you post images online, there's no 100% guaranteed way to prevent this, and you can probably assume that there's no need to remove/edit existing content. you might contest this as a matter of data privacy and workers' rights, but you might also be looking for smaller, more immediate actions to take.

...so I made this list! I can't vouch for the effectiveness of all of these, but I wanted to compile as many options as possible so you can decide what's best for you.

Discouraging data scraping and "opting out"

robots.txt - This is a file placed in a website's home directory to "ask" web crawlers not to access certain parts of a site. If you have your own website, you can edit this yourself, or you can check which crawlers a site disallows by adding /robots.txt at the end of the URL. This article has instructions for blocking some bots that scrape data for AI.

HTML metadata - DeviantArt (i know) has proposed the "noai" and "noimageai" meta tags for opting images out of machine learning datasets, while Mojeek proposed "noml". To use all three, you'd put the following in your webpages' headers:

<meta name="robots" content="noai, noimageai, noml">

Have I Been Trained? - A tool by Spawning to search for images in the LAION-5B and LAION-400M datasets and opt your images and web domain out of future model training. Spawning claims that Stability AI and Hugging Face have agreed to respect these opt-outs. Try searching for usernames!

Kudurru - A tool by Spawning (currently a Wordpress plugin) in closed beta that purportedly blocks/redirects AI scrapers from your website. I don't know much about how this one works.

ai.txt - Similar to robots.txt. A new type of permissions file for AI training proposed by Spawning.

ArtShield Watermarker - Web-based tool to add Stable Diffusion's "invisible watermark" to images, which may cause an image to be recognized as AI-generated and excluded from data scraping and/or model training. Source available on GitHub. Doesn't seem to have updated/posted on social media since last year.

Image processing... things

these are popular now, but there seems to be some confusion regarding the goal of these tools; these aren't meant to "kill" AI art, and they won't affect existing models. they won't magically guarantee full protection, so you probably shouldn't loudly announce that you're using them to try to bait AI users into responding

Glaze - UChicago's tool to add "adversarial noise" to art to disrupt style mimicry. Devs recommend glazing pictures last. Runs on Windows and Mac (Nvidia GPU required)

WebGlaze - Free browser-based Glaze service for those who can't run Glaze locally. Request an invite by following their instructions.

Mist - Another adversarial noise tool, by Psyker Group. Runs on Windows and Linux (Nvidia GPU required) or on web with a Google Colab Notebook.

Nightshade - UChicago's tool to distort AI's recognition of features and "poison" datasets, with the goal of making it inconvenient to use images scraped without consent. The guide recommends that you do not disclose whether your art is nightshaded. Nightshade chooses a tag that's relevant to your image. You should use this word in the image's caption/alt text when you post the image online. This means the alt text will accurately describe what's in the image-- there is no reason to ever write false/mismatched alt text!!! Runs on Windows and Mac (Nvidia GPU required)

Sanative AI - Web-based "anti-AI watermark"-- maybe comparable to Glaze and Mist. I can't find much about this one except that they won a "Responsible AI Challenge" hosted by Mozilla last year.

Just Add A Regular Watermark - It doesn't take a lot of processing power to add a watermark, so why not? Try adding complexities like warping, changes in color/opacity, and blurring to make it more annoying for an AI (or human) to remove. You could even try testing your watermark against an AI watermark remover. (the privacy policy claims that they don't keep or otherwise use your images, but use your own judgment)

given that energy consumption was the focus of some AI art criticism, I'm not sure if the benefits of these GPU-intensive tools outweigh the cost, and I'd like to know more about that. in any case, I thought that people writing alt text/image descriptions more often would've been a neat side effect of Nightshade being used, so I hope to see more of that in the future, at least!

[modrew71extra]

Right, never thought I had to make one of these callouts in my entire time of using social media, but unfortunately due to the seriousness of the subject matter at hand, I have to get this out here to prevent more people being mislead.

Now for a lot of people who follow artist Tumblr might be aware of this one thread circulating in regards to people using Glaze & Nightshade in response to the recent updates made to Tumblr's data services. With the posting in question, @ reachartwork discouraging use of it and presenting an argument for it.

Now unlike most other people, I can tell how off their argument is as it lacked the nuance of how these programs worked and talked about it in a way that came off more enthusiastic, despite their claims of having sympathies people trying to protect themselves.

It's only until I took one little look at their blog and knew what was up, and their later behaviour proved my suspicions.

Now to get this right off the bat, I do apologise for whatever misgendering I exhibited as I didn't notice that about them, nor am I justifying this callout to harass them.

So please don't

Rather, it's to be weary of this user, and how you really shouldn't be sleeping on these programs.

Now as the original featured, it went out in presenting their argument as followed:

As you can notice in the thread, it's providing very little evidence of their findings of its supposed workings without proper sources, while trying to show ways to work around it (again, enthusiastically mind you).

And yet as you notice, they didn't bother providing an alternative way to combat this scalping situation everyone is in, especially with how replies have been of users expressing their grievances over this.

You'll think that maybe you should have added something to help others or worded themselves in a more sympathetic manner?

But that's when I checked into their profile and knew what seems to be up,

They're into this tech.

I knew this was clear propaganda, why would somebody who is invested into generative tech try and discourage methods in protecting others of data scalping?

Like I'm sorry but you can't be somebody who claims to be sympathetic about the whole power imbalance this whole field is causing, yet enthusiastically be into this tech. Those mixture of ideologies just do not match.

There's no such thing as "ethical AI use" for this tech.

Now upon knowing many were falling prey to this nonsense, I had to step in and present my own counter arguments and why they are so wrong. Both programs Q&A (along with demonstrations of it acting as written) thay explains the inner workings of it as thoroughly as possible:

https://nightshade.cs.uchicago.edu/whatis.html

(Paper is in QA)

https://people.cs.uchicago.edu/~ravenben/publications/pdf/glaze-usenix23.pdf

https://twitter.com/zer0int1/status/1749574897179742353

Now comes the part when things get more heated

After I posted my retort, this is the response I got:

A) Again, while I didn't intend disrespect for their gender, "AI-bro" has literally become a catch-all for us who are against the very people who are engaged in this field

B) Since when in my original argument that I did try to villainize the entirety of programmers in general? My wording was superficially against those in that field of ML tech who are for this tech.

C) Trying to spin my own findings with proper context and immaturely handwave it by boiling it down as me going "nuh uh" is making you look childish, the info I provided literally explains the very points they argued.

D) Reason I blocked them straight away is because I knew a lot of these pro-AI tend to be very combative when it comes to criticism, I felt it in my gut that they were going to do so first. And later on other factors of their character proved my point.

Like I'm sorry, you really think me not having a degree of X matter is somehow making me a worst person and that I shouldn't be allowed to voice criticism of something?

Just because somebody is in the field of something, doesn't automatically mean they have the best interests to heart.

And here's my confirmation that like a lot of these AI enthusiasts, they're very sensitive & combative.

Now not too long, I decided to try to reblog my argument on another reblog of the artist @ Kang-Bang as they have a bigger artist presence, while they fortunately did realize what the OP they were quickly blocked upon reblogging my own post:

But it wasn't that confirmed that behaviour.

I found out through a conversation I had with somebody on the server of artist-rights advocate Zakugu Mignon, that this individual had a similar encounter with this user a year ago on Twitter (I'm Hollow btw);

And here's the conversation this person had that led to this similar experience:

https://twitter.com/Acfusi/status/1691261322988527617

Notice a familiar pattern of behaviour?

The unfortunate pattern behaviour that these pro-AI types are once again present.

It's always trying to justify the usage of this succeeding.

Now look, I'm not against the idea of the physical unabled being given the ability to produce creations of their own with the help technology means nor do I think that Artificial Intelligence is inherently bad.

But this generative tech is just doing it all the wrong way.

It's by all intentions & purposes, displace hard working people as cheaply and quickly as they can.

It has unfortunately happened to certain working sectors such as journalism, advertisement and translations as we speak.

Now on the topic of whether or not you use Glaze or Nightshade.

Please don't drop it

Now yes it's not a panchea for the societal problems we're currently facing.

But you shouldn't just leave whatever you post online out in the open without any forms of protection is not the wisest route to take.

As the Q&A already shown, it at least offers some way to ensure a means of sabotage data scalping.

There are still many other creatives and general users utilizing these programs for good reason, and WHY they're desperate for a means of protection.

https://www.tumblr.com/astraskylark/741393628982886400?source=share

https://www.tumblr.com/leahfrog/743484550954598400/theres-also-nightshade-if-you-havent-heard?source=share

https://www.tumblr.com/thetreetopinn/738157011350470656/ill-say-what-ive-said-in-the-past-ai-art-can-be?source=share

https://www.tumblr.com/in-ravenlight/743565614387494913?source=share

https://www.tumblr.com/luimnigh/743036171813273600/what-is-this-about-the-tumblr-staff-wanting-to?source=share

https://www.tumblr.com/taikova/738369881482919936?source=share

Yes, I understand there's some skepticism going around and we're all frightfully confused about what to do in these uncertain times.

But please.

Don't skim out on a solution to offer some forms of protection.

Having a little protection is better than having none.

[kiisaes]

hey so what that anon said about ai training on tumblr is pretty likely true--the rumor isn't completely verified, but a former staff member (user @/jv) has heard that there is a possible deal in progress between automattic (tumblr's parent company) and midjourney

if this is true, there are ways to prevent it! some people at university of chicago have created programs Glaze and Nightshade to help artists prevent their art from being scraped by ai. both programs are free. if you have questions, both programs have an faq that you can find by just looking up "glaze/nightshade uchicago" and there are also several tumblr posts floating around with tips and advice from people who have used them

there's no way around it, it sucks that you have to go to such efforts to protect your art, but i thought it might help to know that you don't have to be completely helpless in the face of this kind of dystopian situation

i heard about this! i'll look into it for sure. i did some brief research a while ago but thanks y'all for reminding me, bc honestly i forgot about it ><

it's just so frustrating that social media sites are sucking generative AI dick in the first place yknow

[kiwinatorwaffles]

remember yall uchicago theglazeproject is offering a web version of glaze/nightshade, webglaze! you need to message them on instagram, twitter, or through email to get an account, but it's so useful for those like me who don't have a beefy computer to run glaze. protect your art. stay safe from generative ai.

[catholicide]

i'm a reformed hater now, so as much as i'm skeptical of generative ai hysteria, i'm just as enthralled by the development of techniques to resist it (nightshade, glaze, etc). somewhere in between the pull towards and push against what is obstensibly a new medium, there's an opportunity to watch people feel their way through theories of art. i've really come to love this. i mean, it's a little arrogant and incurious (we're human, it's the internet lol), but even having the balls to just define art as one thing and not another thing, as if it's so easy, idk, it's just fun to watch. it says a lot about an individual's engagement with art, what they consider valuable about art.

anyways, that figure demonstrating uchicago's data poisoning tool is instantly so evocative for me. it brings to mind vandalized iterations of piss christ. incidental meditations on the opposing philosophies of art. i like it a lot, i like rotating it in my mind. stripping away the more circular, boring elements of ai art discourse, there's lots of room for your brain to play i think.

[prokyon]

Nightshade & Glaze your artworks

Glaze: to prevent AI mimicry of your work https://glaze.cs.uchicago.edu/

Nightshade: to poison AI models https://nightshade.cs.uchicago.edu/

You can do both! Please read the user guides. There is also a web version of Glaze if your computer can't run the program.

According to the UChicago Glaze team, if you're going to use both, use Nightshade first and Glaze second.

Note that both of these should be the last step you take before uploading. Take care of adjusting/cropping/watermarking your images before applying Nightshade and/or Glaze.

Be sure to also opt out of allowing Tumblr to directly share your content to train AI models. Settings > Prevent third-party sharing.

We don't know how much of our content could have already been shared by Tumblr, and as a publicly-viewable site it is of course going to be scraped for training AI models regardless, but we can poison and evade these algorithms with tools like Glaze.

[hoyaheartprince]

!! IMPORTANT !! FOR ARTISTS AGAINST AI !!

Now that tumblr has decided to opt in for ai, i'd like to remind (or if you havent heard of it, introduce) people to nightshade and glaze. Glaze and Nightshade are two separate programs by Glaze at UChicago that are made to trick ai! How Glaze works is by putting a "filter" that defends against art style mimicry, whilst Nightshade works by putting a filter over your artwork that poisons ai

**Do note that: its recommended to use both nightshade AND glaze, as nightshade can't protect your art from style mimicry, and glaze cant poison ai models. however there are some risks and limitations i will go into later into this post

The image below shows how nightshade works

Source:

Whilst Glaze and Nightshade is meant to protect your art, there are some limitations and risks.

Changes made by Nightshade and Glaze is more visible on artworks with flat colors and smooth backgrounds. Nightshade and Glaze are actively working to improve on those fronts. Luckily Nightshade and Glaze has settings that would let you edit the intensity of the filter. (however in the lowest settings, changes are still visible, especially with glaze. trust me i tested it lol)

Nightshade and Glaze is not a long term resolution as ai models improve so in the future, glazed/ nightshaded artworks would not work against newer models. Glaze and nightshade are continually updated to work against the improved models

Theres more info of how nightshade and glaze works from the links below: Glaze:

Nightshade:

https://nightshade.cs.uchicago.edu/whatis.html

[silver-tongued-randomness]

For Artists on Tumble

Because of the stinky stinky AI stuff tumble is adding to our webbed site tm make sure to use either GLAZE AND/OR NIGHTSHADE for anything you upload online to protect it from data scraping (link to the Twitter below)

[merlinart]

This tiny blog celebrates Merlin artists, so I hope you'll forgive the interruption of your regularly scheduled artistry-posting to amplify some information related to the above post from @staff.

So let's say an art blog does not want their creations to be used by AI companies. That art blog has opted out of sharing content with third parties as instructed above but, dear reader, is that enough? This mod doesn't think so.

Quotes of note (emphasis my own):

"Engadget emailed Automattic to ask for comment on the report. The company replied with a published statement, claiming, “We will share only public content that’s hosted on WordPress.com and Tumblr from sites that haven’t opted out.” The statement notes that legal regulations don’t currently require AI companies’ web crawlers to abide by users’ opt-out preferences." [...] So, if a Tumblr or WordPress user requests to opt out of AI training, Automattic will allegedly “ask” and “advocate for” their removal. And the company’s AI boss “believes” the AI companies will find it in their best interest to comply “based on our conversations.” (How’s that for reassurance!)

Staff's commentary on "proposed regulations" in their above post is a false flag of security. The fact remains that AI is a vastly unregulated industry -- if you want to protect your work beyond toggling a setting that is in essence asking politely for companies to respect your preferences, you'll want to employ additional mechanisms.

A defensive option is UChicago's Glaze, an application which, put simply, processes your art to protect against style mimicry (it's worth reading at length about the technology and goals on their site). But have you heard? The same team behind Glaze has launched WebGlaze for artists who may not have the computing power to run the full Glaze application or are primarily mobile device creators.

If going on the offensive is more your style, look into Nightshade. Also from UChicago researchers, this application turns your art into visual belladonna for models who'd consume it. From their site:

Nightshade can help deter model trainers who disregard copyrights, opt-out lists, and do-not-scrape/robots.txt directives. It does not rely on the kindness of model trainers, but instead associates a small incremental price on each piece of data scraped and trained without authorization.

For those of you with extra fire around this issue, consider reaching out to your appropriate government representatives or legal bodies to encourage more action on AI regulation.

♥ you all, @merlinart

Hi, Tumblr. It’s Tumblr. We’re working on some things that we want to share with you. 

AI companies are acquiring content across the internet for a variety of purposes in all sorts of ways. There are currently very few regulations giving individuals control over how their content is used by AI platforms. Proposed regulations around the world, like the European Union’s AI Act, would give individuals more control over whether and how their content is utilized by this emerging technology. We support this right regardless of geographic location, so we’re releasing a toggle to opt out of sharing content from your public blogs with third parties, including AI platforms that use this content for model training. We’re also working with partners to ensure you have as much control as possible regarding what content is used.

Here are the important details:

We already discourage AI crawlers from gathering content from Tumblr and will continue to do so, save for those with which we partner. 

We want to represent all of you on Tumblr and ensure that protections are in place for how your content is used. We are committed to making sure our partners respect those decisions.

To opt out of sharing your public blogs’ content with third parties, visit each of your public blogs’ blog settings via the web interface and toggle on the “Prevent third-party sharing” option. 

For instructions on how to opt out using the latest version of the app, please visit this Help Center doc. 

Please note: If you’ve already chosen to discourage search crawling of your blog in your settings, we’ve automatically enabled the “Prevent third-party sharing” option.

If you have concerns, please read through the Help Center doc linked above and contact us via Support if you still have questions.