#AI information

[sunbentshadows]

Hey all, you know how internet searches suck now? When the results are awful, full-of-AI, death-of-the-internet levels of bad?

Start appending date constraints to your searches - "before:2023".

My results have gone from 90% AI bullshit to ~60% usable - which frankly at this point is a huge improvement.

[8pxl]

and get sued 😤

[rockpapercynic]

There's an epidemic of AI-made photographs on this platform. Right now, they're mostly used for farming clicks, but as A.I. models get better at faking images, be prepared for a spread of scams, misinformation and propaganda.

Here are some tricks for spotting AI!

[starfall-isle]

[eileen-crys]

AI DISTURBANCE "OVERLAYS" DO NOT WORK!

To all the artists and folks who want to protect their art against AI mimicry: all the "AI disturbance" overlays that are circulating online lately DON'T WORK!

Glaze's disturbance (and now the Ibis Paint premium feature, apparently. Not sure.) modifies the image on a code-level, it's not just an overlayed effect but it actually affects the image's data so AI can't really detect and interpret the code within the image. From the Glaze website:

Can't you just apply some filter, compression, blurring, or add some noise to the image to destroy image cloaks? As counterintuitive as this may be, the high level answer is that no simple tools work to destroy the perturbation of these image cloaks. To make sense of this, it helps to first understand that cloaking does not use high-intensity pixels, or rely on bright patterns to distort the image. It is a precisely computed combination of a number of pixels that do not easily stand out to the human eye, but can produce distortion in the AI's “eye.” In our work, we have performed extensive tests showing how robust cloaking is to things like image compression and distortion/noise/masking injection. Another way to think about this is that the cloak is not some brittle watermark that is either seen or not seen. It is a transformation of the image in a dimension that humans do not perceive, but very much in the dimensions that the deep learning model perceive these images. So transformations that rotate, blur, change resolution, crop, etc, do not affect the cloak, just like the same way those operations would not change your perception of what makes a Van Gogh painting "Van Gogh."

Anyone can request a WebGlaze account for FREE, just send an Email or a DM to the official Glaze Project accounts on X and Instagram, they reply within a few days. Be sure to provide a link to your art acc (anywhere) so they know you're an artist.

Please don't be fooled by those colorful and bright overlays to just download and put on your art: it won't work against AI training. Protect your art with REAL Glaze please 🙏🏻 WebGlaze is SUPER FAST, you upload the artwork and they send it back to you within five minutes, and the effect is barely visible!

Official Glaze Project website | Glaze FAQs | about WebGlaze

[zuzsenpai]

I feel like people have been warning about this and I completely ignored it. Deep sigh as I force myself to remember to add “-ai” at the end of every fucking search now

(Is there an extension that turns this off??)

[maruuax]

Target🪶

[habken]

If you dont know already, tumblr is planning a cooperation with Midjourney. Nightshade and glaze your illustrations, if you didn`t already !!

I’m so tired man

[elekilokal]

guys i’ll be honest.. I’ve made it through a lot on here, but if tumblr decides to sell to midjourney or any other generative AI I have to delete everything… 😔 it seems like they’ve already scraped most of the posts on here dating all the way back to 2014, but even if it’s too late to prevent theft of my art, I just can’t keep using a website that would do something like this— I don’t want to contribute clicks and views to a company that so readily throws its artists and creators (the reason anyone uses the site!) under the bus. If it does happen, I will probably make an archival drive of all my posted pieces and upload a nightshaded version to google drive or pixiv or something. If there are any works you really love, I would recommend saving them on your own.

I’ve had my art reuploaded without permission, stolen and sold on random websites, but this generative AI grift is the worst of them all. It’s soulless greed that does nothing but take advantage of those who already struggle to make a living in a notoriously thin job market. It’s a pointless theft machine. As if the world was ever starved for artists.

[gen-z-superheroes]

"Stopncii.org is a free tool designed to support victims of Non-Consensual Intimate Image (NCII) abuse."

"Revenge Porn Helpline is a UK service supporting adults (aged 18+) who are experiencing intimate image abuse, also known as, revenge porn."

"Take It Down (ncmec.org) is for people who have images or videos of themselves nude, partially nude, or in sexually explicit situations taken when they were under the age of 18 that they believe have been or will be shared online."

[yuri-is-online]

tf do you mean Ortho offers to give the player government secrets on their birthday. Why- why does he do that?

[gojoest]

i think satoru can make it through another day so long he keeps believing, wholeheartedly, that he’s the only man you’ve ever been with. keeps his sanity in check so you just play along with it

[kindahoping4forever]

AshtonIrwin: Have you ever woke up from a bad dream? The one where it feels like ya just can’t get anywhere?

[amalasdraws]

It is so important and vital that we all learn how to detect ai and deepfake!! We need to learn media literacy! There is so much ai brain rot going around!! People who fall for clear ai images and fall for propaganda! But also people who claim something is ai when it isn’t just because they don't like what they see.

It is not easy to wade through the huge amount of information we have access to and that is pushed onto us. Ai makes all of this worse and it's dangerous! Be aware and be careful of what you share!! Learn how to detect ai! It will only get better and it will become harder and harder.

[thelaurenshippen]

open ai made a little post about how they chose their chatbot voices and two things stood out to me:

a) I am genuinely pleased and surprised that the voices come from specific real people that they hired and paid, rather than being built off of all of the data open ai has scraped. the voice for siri did one job for one company 20 years ago and now she is literally everywhere without ever being compensated by apple or even acknowledged as the voice. there's simply no way she was paid enough in that original job. whereas, open ai says "each actor receives compensation above top-of-market rates, and this will continue for as long as their voices are used in our products." this could mean literally anything (what market rate are we going above? does continue mean they get residuals or get paid for doing more sessions?) but, christ, at least the actors are doing it with full knowledge of what their voices are being used for and can decide for themselves if the compensation is enough

b) the post talks about working with "award-winning casting directors" to get the voices. first of all, yay for paying casting directors! we love to see it. but they also say the CDs received over 400 submissions in a week and they state that like a big number and it's just...not at all. and look, without knowing the intricacies of the casting process, it's hard to know what approach the CDs took - it's very possible they were selective from the jump and 400 is a lot from the pool they were tapping. 400 would be a lot if you were going to the agencies directly and asking for names, but there's just no way in hell open ai went after big stars for this. so it would've been a pool of unknowns. in which case, 400 is laughably small. even if you're not using the big casting sites like actorsaccess, I've worked on projects with CDs and their own internal systems where we've gotten over 100 submissions for a single role. I've posted roles on casting sites and received literally thousands of submissions in just a few days

look, I have a very limited perspective on this - I am not a casting director (imo, one of the most important and undervalued jobs in hollywood) and I, in fact, hate the process of casting with a passion. but 400 just seemed like such a tiny pool to pull from and, idk, it heartens me! it's heartening to think that there's very little interest from actors and agents to be doing this kind of stuff. and absolutely no shade to the actors who did--I want actors to get their bag however they want as long as it's, like, safe sane and consensual, you know? but there's something encouraging about thinking that open ai hired some big casting agency to get their foot into the voice acting door and people didn't come running

[draconesmundi]

Happy Dracones Monday! The Vishap

Found in the Armenian highlands, in Armenia, Azerbaijan, Iran and Turkey. Here we see one perched on a vishapakar stone ('dragon stone') in Armenia - these stones are often in the shape of fish or have a carving of some sort of animal sacrifice on them (often a bull), and sometimes they have a mix of fish and bull carvings.

This is just my interpretation of a vishap for Dracones Mundi - I chose to go a little more flamboyant with the design rather than make something that looked like a bull or a fish - especially as it's uncertain whether the carvings on the vishapakar are supposed to represent the dragons physically, or if they are more symbolic of summoning good luck for livestock, fishing or fertility.

I post new dragons for my project every Monday on this blog: @draconesmundi