ai-or-not-ai

[ai-or-not-ai]

[ももえ＠〜12/27虎ノ門@mme_daigaku]

[ai-or-not-ai]

this is your periodic reminder that for all the artifacts and errors and "tells" one could possibly list, the only reliable way to actually determine if an image is ai generated is to investigate the source. it is becoming increasingly common for "fake classical paintings" to circulate around curative aesthetic blogs, and everyone should be using this as an opportunity to not only exercise their investigative skills but also appreciate art more in general. you're all checking out the artists you reblog, right? 🫣

so what are some signs to look for? let's use this very good example.

what a lovely late-impressionist piece blended with evocative leyendecker-esque themes! why haven't you ever heard of this artist before? surely tumblr would be all over an artist like this. who is justin brown?

your two options from here are to do a search for the name, or a reverse image search. i prefer reverse image searching, particularly when it comes to a common name like "justin brown". so what does that net?

Immediately, without looking at any text, something is wrong: it barely exists. an actual historical piece would turn up numerous results from websites individually discussing the piece, but no such discussions are taking place. Looking at the text, though, does show the source-- and at least in this case, the creator was honest about their medium.

But let's also look at the "exact matches", in case a source doesn't make itself apparent in the initial sidebar results like this.

This section will often tell you post dates of images, and here it can be seen that the very first iteration of the image was posted 15 days ago. It did not exist online prior to that.

Seeing how long an unsourced image has been floating around is a skill applicable to more than just generative images! See a cool image of an artifact or other intriguing item with a vivid caption? Reverse search it! If all the results are paired with that caption and only go back a few months, you might just have viral facebook spam.

Sometimes generative creators are dishonest about their medium and do not tag it like in the example, so that's when establishing "jpeg provenance" becomes important. While it can be a little trickier to determine if someone is using generative images and not admitting to it if they aren't trying to pass it off as a classic, something to consider is the age of their account and the frequency with which they post. Here are some account red flags:

-Did they only start posting art after 2022, or if they did before, did their style/skill level WILDLY change? Not gradual improvement-- I'm talking amateur graphite portraits straight into complex digital renders. Everyone starts somewhere, newness is not a red flag alone; it's newness combined with existing in a vacuum away from any community.

-Do they post fully-finished paintings several times a week? -Do many of these paintings seem iterative of a similar theme or subject matter ("three well-dressed young men face each other under shade and dappled sunlight")?

-Does their style change in inconsistent ways? An artist that can swap between painting like Drew Struzan and Hokusai should be pretty well known, right? Why is no one hyping this guy?!

-Do they have social media besides the source instagram? If so, what are they posting about? Are there any WIPs? Doodles? Interactions with other artists? Gallery dates? 3am self-doubt posts? Or is it all self-promo? Crypto? Seemingly nothing art-related at all for someone pushing out 3 weekly paintings?

Basically, if it's important to you to omit this stuff when you curate, please don't just smash reblog if the source doesn't seem to be the OP themselves. Seeking out sources was important even before this became an issue, now it is more than ever.

peace n love

[ai-or-not-ai]

this is your periodic reminder that for all the artifacts and errors and "tells" one could possibly list, the only reliable way to actually determine if an image is ai generated is to investigate the source. it is becoming increasingly common for "fake classical paintings" to circulate around curative aesthetic blogs, and everyone should be using this as an opportunity to not only exercise their investigative skills but also appreciate art more in general. you're all checking out the artists you reblog, right? 🫣

so what are some signs to look for? let's use this very good example.

what a lovely late-impressionist piece blended with evocative leyendecker-esque themes! why haven't you ever heard of this artist before? surely tumblr would be all over an artist like this. who is justin brown?

your two options from here are to do a search for the name, or a reverse image search. i prefer reverse image searching, particularly when it comes to a common name like "justin brown". so what does that net?

Immediately, without looking at any text, something is wrong: it barely exists. an actual historical piece would turn up numerous results from websites individually discussing the piece, but no such discussions are taking place. Looking at the text, though, does show the source-- and at least in this case, the creator was honest about their medium.

But let's also look at the "exact matches", in case a source doesn't make itself apparent in the initial sidebar results like this.

This section will often tell you post dates of images, and here it can be seen that the very first iteration of the image was posted 15 days ago. It did not exist online prior to that.

Seeing how long an unsourced image has been floating around is a skill applicable to more than just generative images! See a cool image of an artifact or other intriguing item with a vivid caption? Reverse search it! If all the results are paired with that caption and only go back a few months, you might just have viral facebook spam.

Sometimes generative creators are dishonest about their medium and do not tag it like in the example, so that's when establishing "jpeg provenance" becomes important. While it can be a little trickier to determine if someone is using generative images and not admitting to it if they aren't trying to pass it off as a classic, something to consider is the age of their account and the frequency with which they post. Here are some account red flags:

-Did they only start posting art after 2022, or if they did before, did their style/skill level WILDLY change? Not gradual improvement-- I'm talking amateur graphite portraits straight into complex digital renders. Everyone starts somewhere, newness is not a red flag alone; it's newness combined with existing in a vacuum away from any community.

-Do they post fully-finished paintings several times a week? -Do many of these paintings seem iterative of a similar theme or subject matter ("three well-dressed young men face each other under shade and dappled sunlight")?

-Does their style change in inconsistent ways? An artist that can swap between painting like Drew Struzan and Hokusai should be pretty well known, right? Why is no one hyping this guy?!

-Do they have social media besides the source instagram? If so, what are they posting about? Are there any WIPs? Doodles? Interactions with other artists? Gallery dates? 3am self-doubt posts? Or is it all self-promo? Crypto? Seemingly nothing art-related at all for someone pushing out 3 weekly paintings?

Basically, if it's important to you to omit this stuff when you curate, please don't just smash reblog if the source doesn't seem to be the OP themselves. Seeking out sources was important even before this became an issue, now it is more than ever.

peace n love

[ai-or-not-ai]

It's Time To Investigate SevenArt.ai

sevenart.ai is a website that uses ai to generate images.

Except, that's not all it can do.

It can also overlay ai filters onto images to create the illusion that the algorithm created these images.

And its primary image source is Tumblr.

It scrapes through the site for recent images that are at least 10 days old and has some notes attached to it, as well as copying the tags to make the unsuspecting user think that the post was from a genuine user.

No image is safe. Art, photography, screenshots, you name it.

Initially I thought that these are bots that just repost images from their site as well as bastardizations of pictures across tumblr, until a user by the name of @nataliedecorsair discovered that these "bots" can also block users and restrict replies.

Not only that, but these bots do not procreate and multiply like most bots do. Or at least, they have.

The following are the list of bots that have been found on this very site. Brace yourself. It's gonna be a long one:

@giannaaziz1998blog

@kennedyvietor1978blog

@nikb0mh6bl

@z4uu8shm37

@xguniedhmn

@katherinrubino1958blog

@3neonnightlifenostalgiablog

@cyberneticcreations58blog

@neomasteinbrink1971blog

@etharetherford1958blog

@punxajfqz1

@camicranfill1967blog

@1stellarluminousechoblog

@whwsd1wrof

@bnlvi0rsmj

@steampunkstarshipsafari90blog

@surrealistictechtales17blog

@2steampunksavvysiren37blog

@krispycrowntree

@voucwjryey

@luciaaleem1961blog

@qcmpdwv9ts

@2mplexltw6

@sz1uwxthzi

@laurenesmock1972blog

@rosalinetritsch1992blog

@chereesteinkirchner1950blog

@malindamadaras1996blog

@1cyberneticdreamscapehubblog

@neomasteinbrink1971blog

@neonfuturecityblog

@olindagunner1986blog

@neonnomadnirvanablog

@digitalcyborgquestblog

@freespiritfusionblog

@piacarriveau1990blog

@3technoartisticvisionsblog

@wanderlustwineblissblog

@oyqjfwb9nz

@maryannamarkus1983blog

@lashelldowhower2000blog

@ovibigrqrw

@3neonnightlifenostalgiablog

@ywldujyr6b

@giannaaziz1998blog

@yudacquel1961blog

@neotechcreationsblog

@wildernesswonderquest87blog

@cybertroncosmicflow93blog

@emeldaplessner1996blog

@neuralnetworkgallery78blog

@dunstanrohrich1957blog

@juanitazunino1965blog

@natoshaereaux1970blog

@aienhancedaestheticsblog

@techtrendytreks48blog

@cgvlrktikf

@digitaldimensiondioramablog

@pixelpaintedpanorama91blog

@futuristiccowboyshark

@digitaldreamscapevisionsblog

@janishoppin1950blog

The oldest ones have been created in March, started scraping in June/July, and later additions to the family have been created in July.

So, I have come to the conclusion that these accounts might be run by a combination of bot and human. Cyborg, if you will.

But it still doesn't answer my main question:

Who is running the whole operation?

The site itself gave us zero answers to work with.

No copyright, no link to the engine where the site is being used on, except for the sign in thingy (which I did.)

I gave the site a fake email and a shitty password.

Turns out it doesn't function like most sites that ask for an email and password.

Didn't check the burner email, the password isn't fully dotted and available for the whole world to see, and, and this is the important thing...

My browser didn't detect that this was an email and password thingy.

And there was no log off feature.

This could mean two things.

Either we have a site that doesn't have a functioning email and password database, or that we have a bunch of gullible people throwing their email and password in for people to potentially steal.

I can't confirm or deny these facts, because, again, the site has little to work with.

The code? Generic as all hell.

Tried searching for more information about this site, like the server it's on, or who owned the site, or something. ANYTHING.

Multiple sites pulled me in different directions. One site said it originates in Iceland. Others say its in California or Canada.

Luckily, the server it used was the same. Its powered by Cloudflare.

Unfortunately, I have no idea what to do with any of this information.

If you have any further information about this site, let me know.

Until there is a clear answer, we need to keep doing what we are doing.

Spread the word and report about these cretins.

If they want attention, then they are gonna get the worst attention.

[ai-or-not-ai]

[ai-or-not-ai]

Friendly reminder that “Emile Corsi” is not a real artist but a gimmick to create a fake 19th century artist using AI generated images.

Welder Wings, however, is a wonderful collage artist.

The Green Knight. Artists: Emile Corsi & Welder Wings

Special post: art that reminded me of Victor von Doom because today is the birthday of Julian McMahon, aka Doctor Doom himself

Dear Julian, happy birthday. I love you so much. You will always by my Victor and most importantly, my source of comfort and inspiration ♡ Words can never come close to describe how much you mean to me. I love you always.

[ai-or-not-ai]

there's a popular twitter artist right now who is getting called out for ai on a drawing, because people are circulating screenshots from two years ago when they were excited about the possibilities that AI offered for going back to making their comic again which had slowed to a crawl because of their RSI. only the thing is the drawing they're getting accused of using AI for has no ai use in it at all, and they've needed to publicly publish their patreon exclusive process video to prove this to an internet mob because of the influx of hate they're getting! good job protecting artists, guys!

[ai-or-not-ai]

Confident these are AI generated though I would give it (a teeny tiny amount of) credit for throwing some texture in the center of the petals.

Let’s zoom in and talk about whats going on that makes me say this:

This post is tagged as photography but these flowers look way to illustrative to be photographed. However, this rendering style is not consistent throughout the images either, so I wouldn’t skip to it being illustrated instead but that it was generated with illustrated flowers included in the data somewhere (and also because most people who spend time painting an image don’t want you to mistake it for a photograph for too long).

what are the purple-y textures. is it lavendar? lilacs?? bluebells? It’s not hard to get the distinct shape of those flowers in photography this is just like randomly applied texture.

the right includes the crunchiest jpeg version of (mayyybe) queen annes lace I’ve ever seen.

These look pretty from far away but dont hold up under close scrutiny.

[ai-or-not-ai]

ed zitron, a tech beat reporter, wrote an article about a recent paper that came out from goldman-sachs calling AI, in nicer terms, a grift. it is a really interesting article; hearing criticism from people who are not ignorant of the tech and have no reason to mince words is refreshing. it also brings up points and asks the right questions:

if AI is going to be a trillion dollar investment, what trillion dollar problem is it solving?

what does it mean when people say that AI will "get better"? what does that look like and how would it even be achieved? the article makes a point to debunk talking points about how all tech is misunderstood at first by pointing out that the tech it gets compared to the most, the internet and smartphones, were both created over the course of decades with roadmaps and clear goals. AI does not have this.

the american power grid straight up cannot handle the load required to run AI because it has not been meaningfully developed in decades. how are they going to overcome this hurdle (they aren't)?

people who are losing their jobs to this tech aren't being "replaced". they're just getting a taste of how little their managers care about their craft and how little they think of their consumer base. ai is not capable of replacing humans and there's no indication they ever will because...

all of these models use the same training data so now they're all giving the same wrong answers in the same voice. without massive and i mean EXPONENTIALLY MASSIVE troves of data to work with, they are pretty much as a standstill for any innovation they're imagining in their heads

[ai-or-not-ai]

[ai-or-not-ai]

These are AI generated.

on the left we see an example of what tends to happen when ai images are generated off of original images that had a watermark. It seems like these out of focus dewdrops are trying to form letters (and I can see someone hiding their watermark along an out of focus stem).

on the right I wanted to emphasize how even though the sharpness of the foreground flowers implies they are highly in focus, when you zoom in the shapes all blend together and become indecipherable in a way that would not occur in an in-focus photo.

Next time you see a photoset of flowers that are pastel and dreamy and absolutely COVERED in dew drops, I’d encourage you to take a closer look.

[ai-or-not-ai]

These are AI generated.

There is no such flower as the description. The very first google result for the name is a website debunking its existence.

The closest in name is Lotus Alpinus which has a yellow flower.

The closest in appearance is diphylleia grayi or the “skeleton flower”, which turns transparent when wet. The shape is distinctly different from the ai image.

Alpine lotus leaf flower

[ai-or-not-ai]

Another AI generated painting by the made-up artist Emile Corsi. There was a previous post that elaborated on this, but this artist is a complete invention; he did not exist.

If you ever have questions about the legitimacy of an artist, you can usually just google their name and find a wikipedia article or go ahead and ask me.

one visual tell here is how the edge of the brick goes in front of figure itself by the arm area, but nowhere else. As if the AI couldn’t decide between it being a painting on canvas or a wall. This would be a very unlikely symbolic choice in the style this painting tries to emulate (which is some kind of 19th century academic orientalism?)

"God Anubis" by Emile Corsi, 1877

[ai-or-not-ai]

Portrait of Princesses Xenia and Nina Georgievna of Russia, Philip de László. 1915.

[ai-or-not-ai]

By Mystical Antiquity

[ai-or-not-ai]

[ai-or-not-ai]

Embroidery