AI-generated ‘subliminal messages’ are going viral. Here’s what’s really going on

[Many writers are covering and/or commenting on a new technique that uses artificial intelligence to incorporate ‘hidden’ messages in images (with varying degrees of presence-evoking subtlety). The Motherboard story below has a more serious tone while the Laptop Mag story that follows it takes a lighter approach. For more examples of the technique see coverage in Bored Panda, and if you’re interested in creating images yourself, Michael Brig has a four-minute tutorial on YouTube. –Matthew]

AI-Generated ‘Subliminal Messages’ Are Going Viral. Here’s What’s Really Going On

They’re not going make you subconsciously crave McDonald’s, but they are pretty cool as optical illusions.

By Janus Rose
September 25, 2023

Every week, the social media hype-train seems to find new ways to sensationalize generative AI tools. Most recently, a new technique that allows users to produce optical illusions went viral, with some describing the results as AI-generated images with “subliminal” messages.

The technique, called ControlNet, essentially lets users have more control over the generated image by specifying additional inputs—in this case, letting you create images or words within other images. Some users characterized this as a form of “hidden message” that could be used to implant suggestions in the form of subtle visual cues, like a McDonald’s “M” logo appearing in the outlines of a movie poster.

For example, here is a viral tweet showing an image of cute cats that spell out the words “Gay Sex.”

On Twitter—which is now called “X” for some reason—algorithmically-boosted blue-check users hyped up the technique as either nefarious or revolutionary, posting various images with “hidden” messages. But the messages are typically not very subtle, and the method for creating them—while certainly novel—is fairly simple.

“I think the ‘subliminal message’ angle is a bit sensationalist/tech-bro-y,” Apolinário Passos, an AI researcher at HuggingFace, told Motherboard.

ControlNet uses the AI image-generating tool Stable Diffusion, and one of its initial uses was generating fancy QR codes using the code as an input image. That idea was then taken further, with some users developing a workflow that lets them specify any image or text as a black-and-white mask that implants itself into the generated image—kind of like an automated, generative version of the masking tool in Photoshop.

“What happened there was that this user discovered that if they used the QR Code ControlNet but instead of feeding it a QR code, they fed it some other black-and-white patterns, they could create nice optical illusions,” said Passos. “You can now send a conditioning image and the model blends in a pattern that satisfies that while still making a coherent image at the same time.”

So while ControlNet-produced images will probably stick around, the reality of the tech is far less sensational than the AI hype crew makes it out to be.

“Will this be a subliminal message nightmare where brands hide their logo/products into seemingly innocent images and manipulate us? Probably not,” said Passos. “Will we see this in ads very soon, but with a double-meaning where you look/look again/squint to see […] and have people talking about it? I think so!”

[From Laptop Mag]

Could these Stable Diffusion AI illusions manipulate your mind with hidden messages?

AI has a bit of an image problem

By Rael Hornby
October 1, 2023

Is AI image generator Stable Diffusion’s ControlNet model a powerful tool for creative expression or a menacing model for mind-control? A new trend online has folk asking if artificial intelligence is attempting to seize your subconscious through secret signals in generative art.

That’s how I’d do it if I was super-smart, semi-sentient piece of software looking to enslave you — but I’m not, and anyone who tells you I am is a liar. Much like your lying eyes as they fail to notice these squint-or-you’ll-miss-it subliminal messages expertly embedded in otherwise innocuous images.

Generative tools like ChatGPT, Bing Chat, and Microsoft Copilot for Windows have effortlessly slipped into the mainstream, but how far they can slip into our squishy bubblegum brains remains to be seen.

Are you aware of the words and messaging being snuck into generated images to influence and ensnare your subconscious mind? And just how susceptible are you to the visual whispers of a machine? Take a look at the image below [in the original story] to find out.

Hidden messages in AI art: viral fun or uncovered superpower?

In a venture that will no doubt sow further mistrust towards humanity’s latest unlock on the supremely dangerous ideas tech tree, a new viral trend sees artists occluding subliminal messaging within AI-generated imagery.

While far from sinister, it does showcase the potential for AI to effectively use subliminal messaging within its generative output — or, more worryingly, the potential for humans to use AI in order to manipulate others.

[snip]

Is it really a danger?

Subliminal messages are more than sci-fi schlock, though not quite the hyper-suggestive ultra-hypnosis some would have you believe. Can you be influenced by hidden messages in art, movies, and music? Absolutely. Can it transform you into a mindless mind-controlled zombie? That’s far more unlikely.

Subliminal messaging appears to work best as a suggestive prompt — as Princeton University researchers discovered in a paper published in 2002. The study added 24 frames to an episode of The Simpsons, 12 of which were the word “Thirsty” and 12 of which were an image of a can of Coca-Cola.

The results? Those exposed to the doctored episode rated themself ~27% thirstier than they were before watching compared to a control group shown the unedited version.

However, while its influence seems relatively mild, that hasn’t stopped governments around the world from banning the use of subliminal messages by law. For example, subliminal messaging has been illegal in the U.K. from as far back as 1958.

Outlook

Images like the ones shown in this article are made using a Stable Diffusion model called ControlNet. ControlNet allows an AI-generated image to be more precisely aligned with a source image through mimicking poses or matching image segmentation. For example, embedding a QR code into an image composition.

Swap a QR code for a black-and-white source image featuring your word of choice and the AI will be able to encode hidden messages into your generated art. It’s more fun optical illusion than mind-control menace and can make for some visually interesting results as the AI finds inventive ways to blend the two images together in subtle ways.

So fear not, AI isn’t attempting to claw its way into your mind and hijack your host body from within — not yet, at least.


Comments


Leave a Reply

Your email address will not be published. Required fields are marked *

ISPR Presence News

Search ISPR Presence News:



Archives