top 43 comments

sorted by: hot top controversial new old
[–] 58 points 1 year ago* (last edited 1 year ago) (7 children)

Will Smith has been having a Kanye-esq descent into madness for several years now, so this comes as no surprise.

Edited to fix spelling error.

  • source
  • hideshow 7 child comments
  • [–] 9 points 1 year ago (5 children)
  • [–] 3 points 1 year ago (4 children)
  • [–] 13 points 1 year ago (3 children)

    You also meant "-esque", unless Kanye is a licensed attorney.

  • source
  • parent
  • hideshow 3 child comments
  • [–] 9 points 1 year ago (2 children)

    Y'know, I even looked at it when I wrote it and thought "this doesn't look right", but posted it anyway. I'll leave it to remind myself how stupid I am.

  • source
  • parent
  • hideshow 2 child comments
  • [–] 31 points 1 year ago

    I'm an expert in fake audiences: From Madden 94 to Madden 25, I've seen them all.

    This is a pretty good fake audience: I'll give it a 13. He should try adding Miis. I always like that you could find your friends in Wii games audiences.

  • source
  • [–] 51 points 1 year ago

    Smith: "Hey now, conjoined twins are one of my biggest demographics!"

  • source
  • [–] 35 points 1 year ago (6 children)

    Best way to know for certain is to interview people from the audience.

    Excuse me, sir?
    Excuse me, sir?

    Oh. Oh my. I'm so sorry. You and your friends are obviously busy, I'll just g-

    Oh god, help me! The nightmares are real!!!!

  • source
  • hideshow 6 child comments
  • [–] 19 points 1 year ago (3 children)
  • [–] 43 points 1 year ago (16 children)

    Meanwhile, the woman in front of him is seemingly holding his hand, but the headband of the woman behind her is somehow over her wrist.

    I’ll never understand how a statistical model for “next most probable pixel” can arrive at shit like this.

  • source
  • hideshow 16 child comments
  • [–] 37 points 1 year ago (1 child)

    I'm still in awe how it gets things correct from randomness. Even the early stages years ago that were obviously bad still would get close to the idea, and now it's not that easy to determine some video origins.

    I'm not talking about the morality and ethics of how it got here, but the science of how it works.

  • source
  • parent
  • hideshow 1 child comment
  • [–] 25 points 1 year ago (7 children)
  • [–] 6 points 1 year ago (6 children)

    It looks like she's holding the lower left hand corner of the sign and his hand overlaps hers slightly, that doesn't look generated.

    But that doesn't explain her bracelet slipping off onto the other womans head, or his weirdly shaped arm where his tattoo is by his watch.

  • source
  • parent
  • hideshow 6 child comments
  • [–] 4 points 1 year ago (1 child)

    Try overlapping your fingers the way shown in the picture. Unless there's something special or wrong with my hand, I can only overlap my fingers that way from pinky to pointer. And if you look at the bottom, it looks like there's a thumb wrapping around.

    Which means that's not his right hand, it's his left hand upside down. But we can see that his left hand is on the right side of the sign—it's the one with a pointer that looks like a cross between ET's finger and a penis.

  • source
  • parent
  • hideshow 1 child comment
  • [–] 0 points 1 year ago

    Looks like it could be that one's wearing a bracelet and the other's wearing a headband and the two are just coincidentally lining up.

    Frankly, I think this kind of CSI "zoom and enhance!" analysis is kind of silly generally speaking. Does it matter if these images are AI generated? Does Will Smith matter?

    Maybe this is AI-generated. Could be. Maybe it's real, could also be. None of the "defects" seem particularly egregious. But at the end of the day I don't see how it's going to change anything of any significance either way.

  • source
  • parent
  • [+] -12 points 1 year ago* (2 children)

    I’ll never understand how a statistical model for “next most probable pixel” can arrive at shit like this.

    Probably because you have no experience writing or reading code for AI. I doubt you've read a single book on AI or taken any classes related to it.

    Why would you think you could understand this? Dunning-Kruger effect?

  • source
  • parent
  • hideshow 2 child comments
  • [–] 2 points 1 year ago

    I’ve read enough about how the langauge models work to develop a statistical graph of words based on a huge corpus of training input. But no amount of training pixels adds up to z-index errors like this. Also, wow your attitude is hostile as hell. I feel sorry for you having to walk around like that. Try not to have a terrible day now, hm?

  • source
  • parent
  • [–] 36 points 1 year ago (1 child)
    load more comments (1 reply)
    [–] 9 points 1 year ago (3 children)

    I'm not up to date on celeb news. Is he the guy whose wife shat on his bed?

    Or the other one that slapped someone on stage?

  • source
  • hideshow 3 child comments
  • load more comments (1 reply)
    [–] 14 points 1 year ago

    Out of all the celebs that should be allowed to use AI, Will has had to watch his AI avatar fumble through eating spaghetti more than any person probably wants to.

    Also, he’s cringe so whatever.

  • source
  • [–] 5 points 1 year ago

    Why didn't he at least try to hide it with horrible video compression?

  • source
  • [–] 1 point 1 year ago*

    Are we sure this isn't a video google "enhanced with ai".

    It's not likely, considering he's got enough money to be a problem for them, but it would tie into thier plan to make everything look like slop so you can no longer tell what is slop.

  • source
  • [–] 2 points 1 year ago
    [–] 2 points 1 year ago

    I like the caption of "accused" like we can't be certain that it is or isn't AI.

  • source