Lets assume that it's possible to generate every possible 400x400 jpg in existence. Will I be able to see every possible situation, every possible human that has ever lived or will live? Every piece of art that hasn't been made yet? Will I be able to see taylor swift naked?
How many images will this generate?
Other urls found in this thread:
babelia.libraryofbabel.info
libraryofbabel.info
babelia.libraryofbabel.info
twitter.com
Oh look, it's this thread again.
160000!*32000000 or something like that
If 400x400 jpg is ok for you, then yes.
Someone calculate how many different images there can be, assuming a 256 color palette
Theoretically, you would also have a zoomed in version of every image. So for a 400x400 image, there will be one that is a zoomed in version of only the first 20x20 pixels of the original, does that make sense?
So you could generate an arbitrarily large image from those 400x400 tiles.
You couldn`t create a large image using those tiles, because it`d look like a glitch.
This would be the first image generated, then the first pixel will change to every color/shade and then repeat and repeat until every possibly image be generated.
160000^256 = 10^1300 combinations or so
for comparison, a password often said to be uncrackable like a 40-letter latin alphabet string would be 40^26 = 10^40 or so
Look at the first pixel.
But there is an image of every possible thing, then surely there would be another image of the same thing but closer up
That's more the estimated amount of atoms on Earth.
It`s quite impossible to locate the similar image that complain the puzzle
>Lets assume that it's possible to generate every possible 400x400 jpg in existence
Funny joke.
>How many images will this generate?
Assuming standard 24bpp color and a 400x400 image, there are roughly 10^1100000 possible arrangements.
Super fucking impossible.
you can use somehing bigger than that.
It's so much larger than the atoms in the known universe, that if the number of atoms in the universe as a whole were a number with 1000 extra zeroes on the end, it would still be smaller than this number of possible combinations.
If you have every possible 400x400 permutation, then you know for sure it exists somewhere in the set.
A 400x400 image of a newspaper might not have legible text, but somewhere in the set there is a closeup of the first sentence.
>Will I be able to see taylor swift naked?
!
da erf
>160000^256 = 10^1300 combinations or so
This is wrong. It would be 256^160000, or 2.5*10^385318.
If every proton, neutron, and electron in our universe was actually a universe of it's own, there's still (far) more possible images for a 256 color 400x400 image than there would be protons, neutrons, and electrons in the combined universes.
don't worry, there also exists tileset of every larger image with markers on them to make them easy to stitch together!
Shit, nigger.
Time to make an 400x400 brute force image generator
>if i have a digital representation of every thing ever, will I have a digital representation of this specific thing?
u srs?
you will have the same thing as now, when someone does a shoop so convincing the only way you know its fake is because they tell you. You'll have swift in every position/outfit you can think of, but no way of knowing which images are the analog for an actual picture of her in that position/outfit.
8 colors 8x8 image. 1 image per second
how long would it take to see them all?
you might have some trouble with that
Two hundred trillion trillion trillion trillion years.
Exponential functions
The problem here is you are all assuming you are going to look at the pictures manually lmao
you create a program that selects the pictures that aren't garbage
>(8^8)^8
Holy shit, what a fuckup. Is that the best you can do?
it's certainly possible to generate them all, it's storing them all that's not possible
How about 2 x 2, then? Surely it can't be that much
you'd still have to generate them all, unless you're able to create God to pick them out.
Did you fail out of algebra?
Have you taken algebra?
my bad
How would you evaluate an image to be not-garbage?
That's subjective
Yes you would. However, the number of images you would generate would be so large that it would be impossible to see a fraction of them.
Think larger than the number of atoms in the universe big.
Still 8 colors, 1 image per second? That's much more reasonable Would take about an hour and 15 minutes.
>thinking a resolution this small will ever yield an interesting image
Still wrong. Let's say we have a single pixel. A 1x1 image. How many possibilities for 8 colors? According to you, (1*1)^8 = 1 and only one possibility. Clearly wrong. You have it backwards, which is also giving you faaaaaaaaar lower numbers than the correct answers.
?
That amount of images it would generate.
damnit
>How about 2 x 2, then? Surely it can't be that much
a 2x2 square with alternating black & white images
it would take 16 seconds
>you might have some trouble with that
It would require a few reincarnations.
Makes me think of this: libraryofbabel.info
>library of babel [...] Thus, it would contain every book that ever has been written, and every book that ever could be - including every play, every song, every scientific paper, every legal decision, every constitution, every piece of scripture, and so on. At present it contains all possible pages of 3200 characters, about 104677 books.
How could I write image generator for i.e. 20x20 pixels?
Uh, no. With a 2x2 black and white, there is only 4 possible combinations, when you consider rotating the image.
Here's a stupid question, building off of OP's question, how feasible would it be to build a program that generates the minimum amount of images, generates 1 image each 2 or 3 minutes, and shares the workload?
What I'm thinking is this:
a website that is designed to serve one person for 2 or 3 minutes each and houses the image generator
it generates 1 image within that timeframe
the generator skips boring images (images that are one solid color or the amount of pixels that are unique colors range from 1 to 8000)
each image has a unique string of characters to represent it
the only information that is stored is the currently generated image and it's string
whenever someone generates a new image, the old image gets deleted and the new image is generated based off a string that is the successor (sequentially) to the previous string
After that, it's up to the people using the program to save and share any interesting images they find. The goal would then be not to generate all the images that can exist, but only the ones that are the most interesting.
>mfw I will finally realise my fetish of seeing a 12 yo cosplayer girl strapon-fuck taylor swift while covered in mud
(400*400)^256 which is an insanely high number.
great! you generated constanza!
see? it wasn't so difficult was it?
[spoiler]there's a hidden message in this post[/spoiler]
Sup Forums, everyone
Muh ai machine lurning
I can program a distributed SETI-like generator. If you setup a google-images-like canvas of thumbnails, you can see at a glance if there's anything interesting and zoom in.
It would be a huge fractal noise pattern image with images of things hidden in its pixels.
This shit makes me anxious and panicky.
Yes, but also infinitely more pictures that will never happen. And you can't discern between those categories.
Theoretically this works. Practically there's just no way to generate enough pictures quickly enough and have them be selected automatically for you.
It would take you a lifetime and you'd just sight noise.
Do it.
there aren't enough atoms in the universe to build enough PCs to generate even a fraction of the pictures in a million years.
Except it wouldn't. There's too much noise. It would be almost impossible to stumble upon actual images inside that clusterfuck.
I won't do it unless we find a way to either filter out the noise (image recognition wouldn't work, too many images to filter and it would take too long) or a way to present it to the user so that he can do it quickly.
No you can't.
64x64 at 4 colors is feasible and I'm currently programming it just because it's saturday
Two words.
Quantum computers.
>4^4096
>feasible
making africa a civilized place is more feasible
that means you're not only looking for every variation of a 400x400, but also at all the combinations between those images
the numbers become astronomical
Only 1.0907481 * 10^2466 possibilities, sounds totally feasible. I was going to post the full decimal number but it wouldn't fit in the comment.
enjoy watching tv-like noise for the rest of your life
Instead of generating many images as fast as possible, what about just generating at least a few images every now and then. Provided there is a system in place that influences what images the generator produces (to filter out images that don't have anything recognizable in them), then surely we would see stuff that has never been seen before, right?
When I say generating images based on a string of characters, I mean a string that acts like a set of instructions for how the image should be made. In it's most simplest and least optimized form, it would be like one long string of hex codes that is then parsed by the program to determine what color each pixel should be. Strings that generate "boring" images could be blacklisted from being used by the generator. Boring images would include images that are one solid color or images that are mostly one solid color with a small group or groups of pixels that vary in color.
The stronger the filters are and the more filters there are would then reduce the chances of noisy/static-y images being generated and raise the chances of interesting images (images depicting people, places, or things) being produced.
Consider:
>Generate a random image with dimensions x by x
> Run through a reverse image search
>Record what you find
It would be extremely slow and completely different but an interesting concept.
>there exists a combination of 0 and 1's that's shows a picture of my onitis naked
#!/usr/bin/python
import image_generator
images = image_generator.generate_them_images(20, 20)
>there exists a combination of 0 and 1's that's shows a picture of you shaggin you're mom
Where in the realm of continuous mathematics, which in this case, the answer is yes. Yes you could. Any image represented by tiles will eventually render any sequence of tiles in the correct order given the ability for t -> Inf.
But as for singular images, the answer is no, since there is a limited amount information to express something approaching the infinite (with your phrasing). Not allowing for limited success cases like blurry representations or partial crops.
Very easily with Haar Cascade feature detection with opencv.
how many quadrillion images per nanosecond can you select with that?
(number of different values a single pixel can have)^(400x400)
>mfw this would be possible on a quantum computer
Let's make it a little more feasible by allowing only 1-bit images.
2**(400*400) = 6.2995023e48164
Should be possible in our lifespan, right?
should be done by tomorrow breh
What if you could filter out noise-like images?
E.g. A certain amount of pixels should be roughly the same color as the surrounding pixels
...
It would still take forever because it takes a few microseconds to determine if a given image is too noisy.
It shouldn't be "filtering out", it should be that noisy images aren't generated to begin with. Else you're wasting boat loads of time generating images that don't look like anything. You just need to find a good way to determine if an image is worth generating. In a way, this is what that Google LSD trip generator did, it only generated images that it thought represented an object it had seen before.
we have this thread at least once every month or two, and when someone brings it up, it hangs around for a few days
people just can't comprehend just how big these numbers are
this desu senpai, mods should really start closing this autism
For an AI to filter out noise, it needs to know what isn't noise. So to filter out everything that isn't something like Taylor Swift naked, it needs to know what Taylor Swift naked looks like. And if it knows that, then there is no need for random generation.
it would be far more efficient to train some neural net to draw porn for you
Hey, thats pretty good! I'll start on the logo!
,'^)
It can know what a naked woman looks like and then generate naked women until it generates a naked Taylor Swift
>useing the smylie with de karat noas!
It still will generate more naked women than there are atoms in the universe. It will spend 500 billion years going through all the pictures of naked women looking for taylor swift, assuming something like a 100 a second.
Nice recursion but you forgot your base case.
but there are only a finite number of possible naked women
Same fagging hardcore
What if all existing women right now were to be generated that would already be 3 billion now imagine that in every possible position and ehatever they could possibly be in.(alot)
a lot of those women are going to look alike because of the very low res
But they are going to be different in the eye of the machine
And even if it was just taytay. She would have an almost infinite amount of positions.
Perhaps in the future there will be software which will model what a person would grow up to look like based on their DNA. Then you could just steal one of her hairs.
Enjoy still generating images long after the heat death of the Universe.
I sure hope that returns a generator and not a raw list
You can find each of those images encoded as π digits.