mastodon - Link to source
The media in this post is not displayed to visitors. To view it, please go to the original post.

The Minions of Avian Intelligence

Bonus speedpaint: peppercarrot.com/en/miniFantasyTheater/067.html#bonus

#webcomic #krita #miniFantasyTheater

67. The Minions of Avian Intelligence - Pepper&Carrot

Official homepage of Pepper&Carrot, a free(libre) and open-source webcomic about Pepper, a young witch and her cat, Carrot. They live in a fantasy universe of potions, magic, and creatures.
Pepper&Carrot
#krita #webcomic #minifantasytheater
in reply to Carlos Solís
mastodon - Link to source

David Revoy

for those who wanted to know what the QR code says

Sensitive content

@Carlos Solís @Maaike
in reply to David Revoy
mastodon - Link to source

Netraven

I like the QR codes and "interesting" background dithering. I presume it's to be adversarial to machine training, which is cool. I've been poking around at synthID artifical image provenance watermarking lately and trying to to identify by observing the RGB rasterization layers for any details which might be machine readable. I haven't gotten far but I have some examples of what works and what doesn't. Ultimately I'd like to see if I could import multiple different labs provenance marks on a single image and see how the image models or verification tools respond.
Esta entrada fue editada (ayer, 16:48)
in reply to Toon Link
mastodon - Link to source

Netraven

The media in this post is not displayed to visitors. To view it, please go to the original post.

@ToonLink this is an example of what (little) I know. The green image is the original artifact, the others are images I have merely altered the color palette. All of these will verify are made by ChatGPT. However... if you flip the B&W version vertically, it will no longer verify. One might believe the verifier has a handedness, however, in a more complex image, no amount of flipping or rotating will cause the verification to fail.

Thank you for coming to my TED talk.

@Toon Link
Esta entrada fue editada (ayer, 18:50)
in reply to Toon Link
mastodon - Link to source

Netraven

@ToonLink I am somewhat sure it has something to do with isotropy, just a hunch. Kinda feels like staring at inkblots. The watermark appears to be content aware and can be represented in various different ways, but it must exist measurably on the pixel level of the document. And after identifying it, the watermark doesn’t tell us if there’s any special geometric transform or otherwise which would decode a payload of information like what model made it and when, or if it’s something coarser and simple as a image compression method which reveals which lab made it. I can’t even begin to guess other than whatever I arrive at is still probably wrong.
@Toon Link
Esta entrada fue editada (ayer, 22:15)
in reply to Toon Link
mastodon - Link to source

Netraven

@ToonLink The bad scenario here, as I see it, is someone putting a forged watermark on another person’s art and then claiming the work is AI-generated. That’s why I’m not especially interested in discourse that treats watermark verification as definitive. Even if a verifier is opaque and simply says, “Yes, ChatGPT made this,” the labs might have additional information that could eventually show the watermark was forged. But by then, the representational harm has already been done, because the public has no access to that deeper provenance data. We’re left to infer whatever the labs allow us to infer.

Ultimately, I don’t think obscurity is an acceptable form of security. But I also don’t think publicly disclosing sensitive details and hoping someone else fixes the resulting vulnerabilities is acceptable behavior.

@Toon Link
in reply to Netraven
mastodon - Link to source

Toon Link

@Netraven *scratches chin* Yeah, I think it's one of those things that sound good on the surface, but it has problems. In my case, I worried that it would make people reliant on what they think is sure thing, but it relies entirely on opting in from companies and doesn't help against bad actors who have no interest in transparency or honesty. That doesn't include the sort of problems we've already been seeing with stuff like age verification, which has been twisted for capitalist interests, too.
@Netraven
Esta entrada fue editada (ayer, 22:56)
in reply to Joe
mastodon - Link to source

David Revoy

@Kampfdiestel True. It's really a chance here on Mastodon there is not a lot of AI bots doing AI propaganda. I find a lot of them on Instagram right now. Maybe my comic also act like a honey pot for them. Esp this episode instagram.com/p/Dbqn67rjDQ4/?img_index=1

David Revoy on Instagram: "Authenticity Problem #webcomic #krita #miniFantasyTheater"

75K likes, 781 comments - deevadrevoy on August 5, 2026: "Authenticity Problem #webcomic #krita #miniFantasyTheater".
Instagram
@Joe
in reply to David Revoy
mastodon - Link to source

Gergovie

Il faudrait beaucoup plus de téra$ pour que ça marche vraiment plus beaucoup mieux...
🤮🤮🤮

letemps.ch/cyber/il-n-y-a-pas-assez-d-euphorie-autour-de-l-ia-affirme-openai-alors-que-les-craintes-…

«Il n’y a pas assez d’euphorie autour de l’IA», affirme OpenAI, alors que les craintes d’une bulle augmentent - Le Temps

L’éditeur de ChatGPT demande davantage d’enthousiasme sur l’intelligence artificielle, réfutant tout risque autour des accords avec Nvidia et d’autres géants du secteur. Mais il n’y a pas unanimité sur ce sujet
Anouch Seydtaghia (Le Temps SA)
in reply to Souly🐧
mastodon - Link to source

David Revoy

The media in this post is not displayed to visitors. To view it, please go to the original post.
@Schurke Hey, sure, the simpler is sharing a screenshot of my settings; I keep the same for all my episode weekly peppercarrot.com/0_sources/miniFantasyTheater/making-of/

Index of /0_sources/miniFantasyTheater/making-of

www.peppercarrot.com
@Souly🐧
in reply to David Revoy
mastodon - Link to source

David Revoy

The media in this post is not displayed to visitors. To view it, please go to the original post.
@Schurke And for the recording part; here it is. The comic are in 3840x3840px, and saved at 1920x1920p every 10 seconds. This way, I can also record while working on the laptop; it's almost unnoticeable for the impact on performances.
@Souly🐧
in reply to David Revoy
mastodon - Link to source

Yevheniia

The media in this post is not displayed to visitors. To view it, please go to the original post.

David, ---------!!! PEOPLE !!!--------
PLEASE CONTACT ADRON HALL @adron ASAP. HE IS A DRUG ADDICT, HAS BECOME UNREACHABLE, AND THIS IS A VERY IMPORTANT MATTER. I SINCERELY ASK YOU TO HELP ME CONTACT HIM.

You should be ashamed that, as a U.S. citizen, you cannot even send $42, though you promised to do so two weeks ago. But since you are most likely a drug addict, you apparently do not keep these funds with you. So I will keep trying to reach you so these $42 can be put to good use

@Adron 🤘🏻
in reply to 💡𝚂𝗆𝖺𝗋𝗍𝗆𝖺𝗇 𝙰𝗉𝗉𝗌📱
mastodon - Link to source

Stemy

@SmartmanApps This is not what i'd call a representative enchantillon since most of people doesn't even know the existence of Duckduckgo. And once again, the success of the AI tools and the IA generated content speaks for itself.
@💡𝚂𝗆𝖺𝗋𝗍𝗆𝖺𝗇 𝙰𝗉𝗉𝗌📱
in reply to David Revoy
ActivityPub - Link to source

JaK

if anything cheers me up in the “Age of Avian Intelligence(*)” it is how your comics face the issues with a smile. The fact some friends and some admired bloggers are now enthousiast about this technology is depressing.

Can’t wait to see Cepper in a cyberpunk setting ;-) we know you lurked the genre for a long time. Just one strip, neon lights, some robots and black sunglasses! 😎