Look at the fingers. You always have to look at the fingers
A viral photo of a guy smoking in McDonald's is completely fake — and of course made by AI::Look at the fingers. You always have to look at the fingers
I spend an embarrassingly stupid amount of time on the internet.
Every time I see an article describing something as "viral", it's almost always the first time I'm hearing about it. (Also, coincidentally often the last time)
Nicely done, that's hilarious. The current state of "AI" is deeply embarrassing and massively over-hyped. Maybe a few trillion more dollars of funding will do the trick...
And drinking a refreshing Ç̷̧̡̢̧̧̙͓̮̮͖͎̱̤̳̙̺͍̺̜͇̠̯̗̼̯͍̖͈̝͎͙͓̦̞̝͙̺̼̗̘̲̙̮̺̪̹̪̗͓͍̳̲̼͈̖͗͑̾̈́̓̀͗̈́͜͜͝͝ͅͅơ̸̧̡̢̧̢̨̛̛̛̙̟̦͓̜̯̖̬̩̙͚̭͚̱͓̫͖̙̮̪̗̜̗̻̬͖̝̖͉̩̻̺͇̲̩̦͍͚̝̐̾͆̾̃̓́͋͑̀͑́̆̌̒͂̍̎̊̂̓̆̔̀͗̉̄̍̒̎́̒͊͛͌̑̒̽̿̇̀̅̊̄͆̊̈̿̀̂̓̈́͌͛̃͑͑͊̚̚̚̕͜͝͝͝͝͝ͅk̶̢̨̢̛̛̛̮̻͕̱͙̰̬̟͕̝̪̮̥͎̪̻͇̣̼͚͍̲̪͙̗̹̱̺̆͑͐̈́͛̏̽͐̿́̈́͗̽̏̔̋͊̀̅̽́̀̿̓̉͐͊͊̐̍̆̎̍̀͒̋̈́̔̿̔́͑̽̔́͛̾̂͑̈́̔̈́͌̍̈̄͛̍͐͐̈́̃͌̄̋̈́́̌̓̉̇̿̀̀̚̚͘̚͜͜͜͜͠͝͠ę̶̡̛͚͓̲̼͈̭͚̙̼͕̮̙̫̖͓̜̣͕̠̞̫̗͙̥͚̣̣̲̫̜̟͔̯̫̩̼̤̅͒̌̍̑̀̇͂̄͗̀̊̄̅̔̔͗̂̾̉̓̐͗̋͗͊̀̏̇͛̊̏̃̌͌́̈͋̑̒͒̈͌̈́͑̂̈͋̎̑̓̊̀́̐̚̚͜͝͠͝͝͠͝a̸̢̡̢̧̧̤̜̰̠̩̯͙͔̖̳̳̩̘̥͙͙̙̬̦̺̗̝͛́̍̓̇́̀͆͑͜ć̴̡̨̧̛̖̱̥̙̳͓̤̻̑͐̅̑̏̐͊͒̾̎́͗̋̎̈́͐̌̇̓͆̓̒͐̔̄́̂̊͑̒̉͑͑͂́͛͋͒̊͋̓̓̀͂͒̎͋͋̎̽́͋̀̏͒̒̿̒̽̂̔̂̀̅͂̋͐̋̈̀̿̿́̍̕̚͝͝͠͝͠͝͝͝͠ơ̸̢̨̨̢̲͚̤͖͇̦̯͓͍̤͈̹̖̞͕̗̬̮͔͚͙͎̩̣͓̹͎͚͈̖̚͜l̴̨̢̧̡̢̢̢̡̨̢̧̛̺̳̭̘͙̬͙͖̬̲̠͔͈͎͇̠̼̘̗̭̤̥̖̭̻͎̱͈͓̲̺̬̫̤̻̙̮̬̯̻͖͇͔̹̩̟̰̣͙̫͎̤͔̼̩͙̬̖̟̪̤͇̟̫̭̳͙͈͇̮̞͕̰̮̙̪̳̟͈̿́͌̀̄̽͗̄̄̆͑̔͊̈̊͌̄͒͋̎̓͐͂̀͛̊͊̐͂̀̽̒̂̊̔̆̏̌̍͛̾̉̽̑̾̋̅͛̌̆̎͛͑̉͑̕̕̕̚͜͠͝͠å̸̧̧̨̨̛̛̛͍̖̗̲͙̝̜̲̪͖̟̬̱̜̜͚̳̻͉͎̪̩͎͈̟̻̫̟͖̲̤̜̥̮̝̗̳͖͕̻̞̙̳̤͚̹̳͍̭͔̤̼͍̬̼̌̒̔͂̈͆̆͒͂͌̄͗̌̆̉̀̉̽́͂̔̆͂̽̌̈́̏́̿͋͂̎́͛̑͛̐͗̋̑̀̃̀̇̎͛̊̇̆͊̏͗͂̋̒̎͛͆̅͑̀̿̎̉͊͑̍̚̚͘̕͘͘͜͠͠͝͝͝͝͝ͅͅ
I guess nobody much tried to see if it was fake or not because it was mundane enough to not matter and believable enough because there was a million people like this guy smoking in McDonald's in the 80's. My great grandma was a chain smoker. Literally all the time. I remembe going to McDonald's for lunch in the 80's and sitting there eating while she chews and smokes at the same time.
So there was no reason to suspect or care about it being a real pic.
Looking at the photo knowing it is AI generated, I can definitely see all of the imperfections, however for someone who is just scrolling looking at stuff and not really paying attention, it's easily passable. And I think that's a huge problem, especially with Sora being a thing, you won't be able to believe anything you see on the internet ever again.
I'd even take it further and I'd bet that there are going to be movies made in the future that will brag about how they didn't use AI in the production. This shit is wild
It's a nail in the coffin to know the world wide webs golden age won't have a resurgence and that you'll no longer be able to believe or trust anything you watch, read, or hear, pretty soon.
AI or not, this looks unnervingly accurate enough to be one of my uncles in the 80s, including the open chest denim vest sewn onto a tshirt. The 80s was wild, kids.
It's super accurate all the way down to the C̴̛͔̫̘̳̜̗̹̮͍͉̲̟̰̐̀̅͒͘͜͠o̶͔̙͙̦͎̩͛̓̔̀͋͜ç̸̧̛̯̰̱̤͔͍͖̥͉̟͍̌̄͗̎̅͌̒̀̀͊̈́̓͘͝a̴̳͇͉͌̔̔͋̍͋̏̈́͝c̷̛͎̙̟͘o̴͕̟̳̺͈̤͓̪̟͕̣̮̳͖̦̓́͂̓͗͊̅͠l̷̝̦̟̱̭̗͇̘̳̗̫͐̑͋a̸̡̡̺̹͉̠̺̹͉̙̮̍̏̂̚͝͝
That's not hair growing through the shirt that's simply a bare chest. AI can paint both shirt and chest just fine, but doesn't understand garments, doesn't understand that if you have t-shirt sleeves you don't have a bare chest. Could happen with a dress shirt but then the collar is missing: It's denim-on-skin in the chest and neck area which the AI can easily make consistent because it's continuous, then shirt-and-denim on the shoulders, which also makes sense. It's the combination that doesn't.
But they look normal? People keep saying that, but except the weird position which it's a picture which it could have caught you in a weird position and that's it.
Like we only see 4 finger on one hand but the thumb could be bended and not visible from the angle.
If you zoom yeah...one finger seems to end up converting to burger but that requires quite a deep look.
Also, as was pointed out in the original post a few days ago, the booth in the left part of the background makes no sense in relationship with both the table, and the guy in the right part of the background. Those tables wouldn't be laid out like that. Also guy in the right background has a capcha face, and the person in the way back has no face. No ashtrays, they had brown glass or yellow tin ashtrays. Also, what's up with his shirt? Sleeves, but no collar, and apparently no chest piece.
I saw those things, and yet, sadly, my brain was still trying to make sense of it all, rather than saying "yo, Bear, this image is all kinds of fucked up, probably not real".
And I'm a cynical, suspicious bastard. We're fucking doomed.
I mean it’s super obviously AI generated, what with the yellow fry box with white sides, and red coke cup with gibberish on it. He has a second white cup behind his hands seemingly floating in mid-air, and his hands are together holding… nothing. There’s a sign in the background that seems to say “Modlidawi” instead of “McDonald’s”
Article: "If you take a closer look at the photo..."
Buddy, you don't even have to look that close. The hat of the dude in the background draws your eye immediately, quickly followed by the fries and all the shit going on on that table.
The average person has exactly zero skill at looking at picture details, it's a thing you develop as an artist and then need to consciously break by flipping the image, tilting your head etc. to look at the whole composition again.
The average artist also won't notice when a cow's horns are above their ears (instead of above in front) because they haven't seen a reference cow in ages, and never specialised in drawing (or whatever) cows. The average person without artistic eye does not even have the impulse to zoom in on that kind of thing and literally won't see the little brown hat you put there, you'd have to make it pink for them to ever notice.
Honestly, I don't see anything really wrong with that hat. The back rim looks a bit strange but that might just be because it's a strange hat, and I'm certainly not fashion-conscious enough to spot that it's an anachronism.
The hands and writing are clearly messed up and pretty much the first thing you should have a look at when figuring out whether something is AI. If this wasn't a "photograph" though but an illustration you might also look at a shitty or lazy human artist in many cases. In this case there's a dead AI giveaway as humans would never make that kind of mistake: The shirt. T-shirt sleeves but bare chest? How does that work? Shirt and vest and chest are painted very well, the AI did a good job at the details, thing is it doesn't understand how garments actually work and that the combination of details doesn't make sense.
Or, in short: (current) AIs are quite good at overall composition and lighting, may or may not have their issues with smaller details depending on the detail, and have lots of trouble enforcing consistency on details that aren't directly adjacent to each other. You'll also see things like e.g. the lines of a dunno column being discontinuous at the spots where it moves from background to invisible (because foreground before it).
Many models have gotten quite good at hands (especially SDXL), unless you're trying to do something complicated with fingers grasping into each other and stuff it'll probably get it right. But the list of these things is endless, you'll e.g. see PC keyboards with seven rows of keys. So look at what you know you know (and I don't know jackshit about hat fashion), and check it.
On another note: It's good that the people putting out those kinds of fakes seem to have zero artistic skill. Without AI, with good ole photo manipulation by an actual artist, spotting a fake could be close to impossible. Generating flawless photographs with current AI is also a crapshoot, you'll end up sifting through a gazillion of possibilities, illustrations are way easier because you can explain things away by "yeah it doesn't need to be accurate in that detail, just give off the right impression". If you want a convincing fake you'll still need an actual artist manipulating actual pixels to fix up the AI mistakes.