https://babelia.libraryofbabel.info
This website contains all posible images. There's an image of you having sex with an alien from a distant galaxy, and there's an image of you in the exact moment you die.
If you use a software that can recognize complex shapes, say, a face, how much computer power would you need to find a meaningful image in a reasonable amount of time?
Obviously, doing it manually would take more time than what the Universe will last.
But could automatization reduce it significantly? Or is it hopeless?
Help, /sci/, these numbers are orders of magnitude bigger that I can't comprehend.
>>9154165
>find a specific instance in an infinite unordered set.
Sounds like something you'd want a quantum computer for, but classically this isn't something a computer will help you with.
>>9154165
>>meaningful image
define meaningful. With things like generative adversarial networks, you can take in a bunch of images and generate new images that have the same 'style' as the input images
>>9154165
It doesn't contain them all. It POTENTIALLY contains them all. Imagine the storage space needed if every possible combination was created and stored.
>>9154165
1/ it only contains images with a specific resolution
2/ it also contains a lot of image meaningful, but that won't happn (e.g. all the way you will not die, at all the time that won't happen)
It's hopeless. We barely have a sense of what "meaningful" is. Let alone figuring out which one will be "relevant". And all of these are so disolved in the noisy ones...
If you want clear image, that your brain can process, but are not meant to be relevant, ask an artist.
>>9154165
Finding a specific image in a collection of all possible images is equivalent to just creating it from scratch
>>9154609
That's one sexy statement right there.
>>9154609
Borges, is that you?
>>9154165
Well, I bet that it's actually doable if we cap the total resolution of the images stored and the color range.
Let's say we have an infinite library of 300x300 images with a 256 color range, how many different images are possible with that?
>>9154165
You must sectionalize the areas somehow as you scan them.
Assuming the sectionalizing method is still based on a unit of use to you while the pictures remain of the correct "size" if you will, then a face from any angle can be calculated based on a few charistaristics that it is actively looking out for.
However this is all based on what direction the scan is going and relies heavily on the quantity of scanners and information linkage that eliminates all the similar or duplicate images from the sensors.
In short.
Area ^ Nth Scanners of equal magnitude divided by the source picture.
However there is a trick you can use in topology to quicken the scanners into finding the image faster once it is established what exactly it is you are looking for by launching the scan from the center and touching the edge. Then radialy scanning the square using a complex theorem which I am too drunk to discuss right now.
Using the possible area based on the smallest acceptible parameter unique to a person at any angle you can write a program that searches for this and only this.
Ergo finding yourself 9/10 times in the one picture (depending on the parameter) faster.
In other words ...
for all intensive purposes a dog is a sphere.
>>9154609
it contains all images from the future too
What if you find an image from the future, and then create a copy of that image at some time in the future. Would you be creating that image from scratch, or did the library create it?