From this morning’s YT feed. A really good perspective on AI.
Listen to Adam’s first question & answer, and then replace Walmart with AI. Rest of the logic totally applies. AI output is the black fence, until it needs to get painted white…
Sometimes you need to let the world solve these problems to quote his take away. And maybe that’s a good perspective for creatives in this overall mess.
Oh, come on! It obviously says “Itfallingrami codmoxzz”
I think they are archeopteryx.
Pterodactyls!
Is that the AI or the VFX version of Lorem Ipsum?
These frames look like a live action Mad Magazine. It’ll work for making unsettling images maybe a horror film or something but will it make people want to buy stuff?
I think the only reason anyone was ok with this was because it’s still worth a few clicks just to say “We made our new spot entirely with AI!” I mean it’s the only reason any of us watched it.
The problem AI has it that is gonna be click-worthy for maybe a couple of months more. After that it’s going to need to not look like garbage and, as I think you were also saying, they don’t have that much time. I mean if any of us had tried delivering this to an agency we’d get crucified.
I got as far as 20 seconds - couldn’t face the next seven minutes.
Back to 1990s wallpaper sci-fi with humans, voices and music…
This afternoon’s treat will be 1930s wallpaper sci-fi…
While technically a bit more cohesive, still just an animated neon slideshow for the most part.
This comment was actually the most meaningful:
Yeah, it looks beautiful and all but still a glorified animated slideshow generator. Still no generation of any sets of meaningful actions.
I want to see dudes coming out of their houses and locking the doors behind them, people feeding birds in parks, a girl playing a full round of hopscotch, someone stopping a taxi and jumping in.
Even something as simple as someone grabbing a bottle of water, opening it, drinking from it and putting it back down, Sora or any video model can’t do that…
Those AIs can’t handle any generations that involves anything that requires more that 2 or 3 steps in even minimalistic thought processes and scaling probably won’t fix that since those are things that actually require video generation models to have some kind of foundation that would, if you really think about it, need to be at least as smart as an AGI…
You can’t wing it without some kind of agentic and feedback loop driven generation process.
At least it wasn’t the Pillsbury Doughboy face…
“Gymnastics is the Turing Test for video generation models”
No notes.
Fix the phone number and it’s perfect.
Prompt on Dream Machine: Astronaut dancing in zero gravity holding a gay flag.
It would be a fun art project to train a set of prompts on nothing but porn and funny cat videos.
