And on https://www.aiornot.com/ I got that both images were made by a human and not AI (with the option to say if its correct or not). So it looks like that AI model has been trained to say BOTH images are human when we know that's not the case.
Thanks for double checking on that! I saw someone mentioned cutting out the pixilated part of the picture and then it changes the results too. It's getting crazy.
Alleged original image used in AI photo
https://i.4cdn.org/pol/1697168323785759.jpg
I put that one in, compared to the Shapiro one. The detector I used gave 52H/48AI for the puppy and an 86% human/ 14% AI on Shapiro's.
I cropped both of them to just the actual picture.
https://huggingface.co/spaces/umm-maybe/AI-image-detector
That's the site I found just doing a search.
Obviously, one or both are fake. How are we supposed to deal with this stuff?
Catsfive is using Optic Ai or Not
https://www.aiornot.com/
It then becomes a matter of which one is the better GANs model, on top of watching out for this stuff.
The hugging face AI image detector only has 76 likes on it. It doesn't look heavily used like https://www.aiornot.com/
The less used and less trained an AI image detector is, the less like it's correct.
I also checked and got the same result as you on https://huggingface.co/spaces/umm-maybe/AI-image-detector
https://files.catbox.moe/7rk3hx.png
https://files.catbox.moe/uzo809.png
And on https://www.aiornot.com/ I got that both images were made by a human and not AI (with the option to say if its correct or not). So it looks like that AI model has been trained to say BOTH images are human when we know that's not the case.
https://files.catbox.moe/4udgxc.png
https://files.catbox.moe/njdlnc.png
Thanks for double checking on that! I saw someone mentioned cutting out the pixilated part of the picture and then it changes the results too. It's getting crazy.