I mean, it could be a manual photoshop job. Just because it’s not AI doesn’t mean it’s real.

But also the detector is probably wrong - it’s likely an AI image using a different model than the detector was trained to detect.

…you know people made fake pictures before image generation, right?

"We investigated ourselves and found nothing wrong."

Ignorant Americans, never even heard of the common snailcat

I have two of those cats. I still can't catch them when its time to go to bed.

No spooky eyes, no extra limbs, no eery smile? - 100% real, genuine photograph! 👍

It's a snat. They are not easy to catch, because they are fast. Also, they never land on their shell.

cute snat...

I wanted to get a cat but I discovered I was allergic to the slime trail.

Such a cute kitty snail! Can you post just the picture?

That's a normal housecat. Not sure what people are confused about

Chat is the picture real?

Well duh it detects AI generated images that are at scale and that snail cat is way too small for it

Clearly AI is on the verge of taking over the world.

My guess is the AI was trained on a combination of cat videos and sponge bob.

That is a weird looking rabbit

It has been 0 days since classified military gene research has been leaked by interrogating ai detecting models

honestly, its pretty good, and it still works if I use a lower resolution screenshot without metadata (I haven't tried adding noise, or overlaying something else but those might break it). This is pixelwave, not midjourney though.

Honestly, they should fight fire with fire? Another vision model (like Qwen VL) would catch this

You can ask it "does this image seem fake?" and it would look at it, reason something out and conclude it's fake, instead of... I dunno, looking for smaller patters or whatever their internal model does?

There are a bunch of reasons why this could happen. First, it's possible to "attack" some simpler image classification models; if you get a large enough sample of their outputs, you can mathematically derive a way to process any image such that it won't be correctly identified. There have also been reports that even simpler processing, such as blending a real photo of a wall with a synthetic image at very low percent, can trip up detectors that haven't been trained to be more discerning. But it's all in how you construct the training dataset, and I don't think any of this is a good enough reason to give up on using machine learning for synthetic media detection in general; in fact this example gives me the idea of using autogenerated captions as an additional input to the classification model. The challenge there, as in general, is trying to keep such a model from assuming that all anime is synthetic, since "AI artists" seem to be overly focused on anime and related styles...

So only 2% "not likely to be AI-generated or deepfake"... that means that it's almost definitely AI, got it!? :-P

What tool is this? I assume the tool is also AI based?

So the prophecy is fulfilled

You have never seen the all elusive catsnail? Bummer you should look harder.