in theory AI could recreate every picture in history with just 250x250 pixels
what stopping someone to do this?
Thalidomide Vintage Ad Shirt $22.14 |
CRIME Shirt $21.68 |
Thalidomide Vintage Ad Shirt $22.14 |
in theory AI could recreate every picture in history with just 250x250 pixels
what stopping someone to do this?
Thalidomide Vintage Ad Shirt $22.14 |
CRIME Shirt $21.68 |
Thalidomide Vintage Ad Shirt $22.14 |
the math
What about it?
It's all guesswork. Same as if a human did it.
Interesting, but what's the usecase for this?
t. project manager
an RNG could do this too, or a monkey with a keyboard that just has 1 and 0
there's nothing significant about it
>https://babelia.libraryofbabel.info/about.html
You zoomers are moronic and late to everything as usual. There's no point in actually "creating" them since they're 99.99999% random noise.
so, by containing every combination of colored pixels in a certain resolution, this library contains every single legal and illegal image/video frame, everything that can be possibly seen by eyes
what are the implications of this?
>https://www.rottentomatoes.com/tv/devs
AI is still just guessing what combination is the "real" one, and humans could never go through every image themselves.
https://ansuz.sooke.bc.ca/entry/23
i aint reading all that shit homie
>form my opinions for me
>but ONLY opinions diluted to fit within half a minute of reading time!!!
Just like the "library of babel" contains the exact way you will die...wow...spoooky
The implications are nothing, this is essentially a rhetorical trick. People simply can't comprehend the scale of these things, so they imagine it means something that "it's in there somewhere", when you can just bash your head to the keyboard and get the same result.
To resolve all pictures with just 8x8 resolution in black and white would take 2^64 combinations.
Why would you need "AI" to do that, all you're doing is rgb(0, 0, 0) through rgb(255, 255, 255)
because AI could be "smart enough" to skip gibberish pictures
Then it's the exact same thing as asking AI "create a picture of the first time a monkey evolved" or whatever now, it's just guessing.
but this way you can never recreate the exact copy of the Mona Lisa
this is literally how diffusion models work
ask you shitty diffusion to recreate a famous painting pixel perfect then
Why would you want to skip "gibberish pictures" if your goal is to recreate every picture in history?
the same that prevents LLMs from being coherent at 7b parameters I guess.
A Babylonian Library of images?
>what stopping someone to do this?
the law, some of the images would be illegal
It couldn't recreate my dick
it literally would
>Why not?
Because it would take longer than the entirely lifetime of the universe to generate them all, without even sorting through them.
>AI filter
Assume they are all generated. It would take longer than the entire lifetime of the universe to go through them and decide which ones were actual pictures of something, maybe.
>AI magic
AI can't tell if they are interesting before they are generated. You have to generate them, then the AI looks at them, and decides. There is no way to tell beforehand.
>AI magic!!!
Suppose you have all the images that are actually pictures of something. How do you pick one? How do you specify the one you want to look at?
>AI magic
Suppose they are already in keyword categories. So you enter one keyword. The results are too large to fit in the universe. You enter ten keywords. It will take until the end of the universe to complete the search and return the results.
>AI magic?!
You see here a picture of Tayler Swift's coochie. It smells like sausage.
>Magic?
Go pay some nerd to draw it for you. I ain't doin' that shit.
>You see here a picture of Tayler Swift's coochie
Not seeing it. Where?
I used to think of this to
Somebody beat me to
https://babelia.libraryofbabel.info/