Technology

Yeah, one of them definitely had the fake, post-processed bokeh effect added to it that a lot of phones with "portrait mode" use. Which, to be completely fair, makes that technically an AI-generated image.

I was looking for artifacts of AI generation, and I found them, but I'm still wrong. I can't win.

Yeah what the fuck was that about? Clearly rigged for sensationalism.

the real photos just looked like professional photography, that's how your photos will look like when you hire a photographer

9/10.

AI sucks at reflections, so pay attention to the pupils to see if both eyes reflect similar shapes.

Also a few had odd lines where their neck was like surgically reattached to another body.

Modern phone cameras are passing pretty much all photos through AI filters now to add detail or upscale, so filtered photos are more the norm than exception now, so that doesn't really help people to filter out AI images. Point still stands that it's already gotten extremely good and now requires recognizing tiny details. It doesn't have to go far to get to the point where it'll be next to impossible to tell.

Yep, same with 8/10. Real faces are less symmetrical than AI

Faces are very well done these days, you can go to https://thispersondoesnotexist.com/ and keep refreshing images. Eventuall you'll notice little things, like earrings not matching or sunglasses being 2 different halves melded together, hats and hair can look weird, etc

But most of them appear very human

For me it was the eyes. Pretty much all of the real faces had realistic reflections in the eyes. The AI ones didn’t.

I got 8/10. The blond AI and the dude with glasses were the two I missed.

2/10 here as well. I went fast to test myself.

I've done a lot of stuff with generative models called GANs, like StyleGAN (2?) which I believe these pictures to be from. My main focus is the "hair bubble effect". This works best for people with longer hair, which is why I had 30% wrong in this test. Basically, by starting at images generated by these models for a long time, I started noticing that it is bad at creating the few loose hairs that stand out from the main pack. These plucks of hair often seem to go around some invisible "bubble" or weirdly flow together with the background. So my main point of focus is often the transition between hair and background, or just the hair in general, since that's where it's most likely to mess up. But the images picked here were also intentionally picked to be the most confusing according to the rest of the article, so it's not that weird that these are hard to classify. Some of the real ones looked extremely AI to me, and it was only after the first false positive that I got a lot more careful with labeling some as "AI" than I normally would.

Example, the strands of hair here (though admittedly the effect is not very convincing here):

The eyes. Look for non-circular pupils or noticeably different-sized pupils.

Imagine the data they will collect from the dozens of lemmy users.

Our words might be scrapped, but it won’t help with facial generation.

I think they purposefully picked real people with slightly odd proportions to try and throw you off, but did it a little too consistently making it easier to guess. I think with more work put into it it would be very hard to tell. It's hard to tell and this is just a quickly thrown together blog quiz.

So I figured out the difference here, and it’s gamed to make you think they’re backward.

The real photos appear to be professionally edited with the light, filters, and bokeh being perfect while the AI generated are more like candids you’d take with your phone. This is an intentional move by the author to make this scarier than it actually is, imo. Next to each other we expect the AI to be more ‘perfect’ than the real photos. Once I figured that out I got them all right.

If they’d put candids next to these AI generated images I’m willing to bet you’d have done better.

Same, so no better than random chance

I got the same score, though I got the dark haired woman. I don't remember the one I missed instead.

I feel like trying to identify unedited images and AI isn't as hard, but when they're edited it gives many of the same details that AI generated ones have. Mostly it's the smoothing out of details. This could be just that the AI was trained on a dataset pulled from images available online (which are almost all edited to some extent) or just because the way it generates it makes details not quite clear. Probably a little of both.

And those small tiny issues can be fixed with a Lora or post processing to refine those specific issues that the general model missed.

Same. We're as good as a coin toss... 😭