every image imaginable (and some that aren't)

[batmanmg]

ok your computer screen has a ton of little dots with a different color given to each dot. (well set of 3 dots in different shades) either way...

well what if you set up something that ran through every possible combination of colors for each dot, starting form total black. (all the lights are out) to all white (all the lights are full blast on) since this would take really long and we want it to do this very fast (as fast as alowed by physics) the only way to make this interesting is to print out every single combination... essentialy you'll have an image of everything imaginable by the end of it.

i know... its alot of images, alot of images that look almost exactly alike, and yet soo many more that are compleletly different... but still every future image you'll ever see, if you were to take a digital picture of it and put it on the computer screen, you'll be able to find it in that stack. everything you've ever seen is also in there, every object from every angle, ever page of every book, ever written/to be written, EVERYTHING.

if you have anything to say about this i'd love to hear it... personaly i think it would be mind blowing that if my entire life were video taped, i'd be able to find every frame of it in that pile.

does anyone know how many images it would be?

and any comments on how to make this idea more useful / practical, than sifting through trillions of useless static to find one glimmer of an actual immage, or billions of duplicates, by hand.

or any comments at all?

[bo198214]

Let k be the possible intensities of one point. And let n be the number of points on the screen. Then you have kⁿ possible images on that screen.

For example take usual 1024x768 and assume 16 bit color = 2¹⁶ colors for each point. Then you have
2^16x1024x768=2^12582912 images.
But maybe youd better start the calculation based on the receptors on the retina.

[jinydu]

essentialy you'll have an image of everything imaginable by the end of it.

This isn't specific to computer images. The same could be said of say, a sufficiently skilled painter.

[PWrong]

Except a painter doesn't use pixels, so you could have infinitely many paintings.

[bo198214]

Ifnt everything is quantisized.

[blazes816]

But, you'd also have every book every written as well, a picture of everyone to be, a picture from everyone it the past. That would be so cool.

[moonlord]

... usual 1024x768 ...

That is a dangerous statement to make .

[houserichichi]

If you really break this down to its components are you not actually measuring the total number of quantum states of an inordinately large number of particles being displayed on what we macroscopically call a computer monitor? If so then I think the problem takes a turn for a worse and I'd rather sip hot chocolate and let you guys work it out. :wink:

I'll be the only sane one left in the room.

[bo198214]

But, you'd also have every book every written as well, a picture of everyone to be, a picture from everyone it the past. That would be so cool.

A whole book on a computer monitor? Ok, each page of a book.
But its absolutely not cool.
The majority of these pictures is crap. You really have to search until you hit something meaningful.
For example if you would search for a certain text passage of your favourite book. You have to remember how exactly it was because every variant is also painted. You see you can not even use this to memorize things.

Takes a sip hot chocolate too.

[batmanmg]

well what if we had a network of supercomputers, sift through EVERYTHING, and sort it all out... get rid of all the totaly mess images, becuase you'll get alot thats just static, organize all the images who's shapes are indistinguishably simmilar,

shift out texts that are totaly unlegable bits, masses of jumbled letters, or parts of letters, or parts of words, in other words, if its not legable text, it gets put in the illegible pile.. uh oh... this massive collection will also have every term paper you're ever going to write in exactly your handwriting... creepy... either way...

Edit by Rob: comment removed

using a search engine to find something your looking for will be a pain though. eg: i search for all text pages with physics, i'll get a billion (more than that but im using a billion to mean A LOT) pages of physics stuff that makes no sense even if the computer sifts out whats gramaticaly correct. cuz i'll get things like. newtons law of physiscs states that butter spread on a commet cannot turn left.

so using this idea for information would be near useless. a writer could looks through random texts pages for inspirarion though. or to take your mind of of things.. it will have to be mainly for entertainment purposes.


this is how the idea for all this came to me.. since there are soo many variations you can come up with in chess, if you get every combination of arangements of peices and sort them to what can possibly come after what . and connect it all through a web, then you'll have every game conceivable in front of you. after you play a game you could search through this mass of games and find yours. i applied the same idea to images and a computer screen.

oh yeah and

2^12582912 how many digits is that? my calculator wouldn't let me do it.


oh yeah and anyone think they could work out how long this would take?,

[Keiji]

I had an idea.

To make this process easier, first we create a set of pictures in 4 shades of grey, size 128x128. That is, 65536 pictures.

Then we decide which pictures actually look like something. We take the ones that look like something, and color in each "part" (definition of a "part" is to the discretion of the editor) in various colors. That makes p^c possibilities, where p is the number of parts and c is the number of colors, for each image.

Then we decide which of those possibilites looks nice. We take that image, and enlarge it to 1152x1152 using hq9x. Then we have just created a very nice picture from nothing!

Of course, if this was to be done in practice, we should have Internet software for many people to decide on which images looked nice, and to determine the various parts of an image.

By the way, I moved this to General, seeing as it is a mathematical topic.

[houserichichi]

oh yeah and
2^12582912
how many digits is that? my calculator wouldn't let me do it.

oh yeah and anyone think they could work out how long this would take?,

There are 3787833 digits in 2^12582912 (base 10).

Number of digits = n in base 10.
n = Floor[12,582,912 Log_10(2)] + 1
=3,787,833

Assuming I didn't mess up the calculations, but if we were to go ahead and count off one possible combination of that number every second it would take us around 10^(3787833) seconds to count all possible combinations off.

Since there are 3,787,833 digits and there are 10 possible numbers that each digit can be (0,1,2,3,...,9), there are 10*10*10*...*10 a total of 3,787,833 times which translates to 10^3787833 and we're doing one digit per second so it takes that many seconds

Given that the universe is only something on the order of 13 billion years old I'd say you have quite the feat ahead of you.

The NASA WMAP people say that the universe is between 12 and 14 billion years old so I just averaged it to 13. You choose any number you like, however

13 billion years translates to an 18 digit number (in seconds), just for comparison.

13 billion years * 365 days per year * 24 hours per day * 60 minutes per hour * 60 seconds per minute* = 473040000000000000, an 18 digit number

[Keiji]

oh yeah and
2^12582912
how many digits is that? my calculator wouldn't let me do it.

oh yeah and anyone think they could work out how long this would take?,

There are 3787833 digits in 2^12582912 (base 10). Assuming I didn't mess up the calculations, but if we were to go ahead and count off one possible combination of that number every second it would take us around 10^(3787833) seconds to count all possible combinations off. Given that the universe is only something on the order of 13 billion years old I'd say you have quite the feat ahead of you.

13 billion years translates to an 18 digit number (in seconds), just for comparison.

And what if one of the six billion people in the world counts off one image every second?

... Okay, it wouldn't reduce it by much. I still like my method, though.

[batmanmg]

well im not talking about counting it, im talking about printing it... or at least storing it in a database, to be sifted through by a network of supercomputers. (anyone know about how fast they'd be able to go through these images)

so maybe instead of 1 per second. lets say we get a computer that can do 10 billion per second.. but like rob said that won't cut down the time enough... so we get a group of a 100 of these computers and separate the possibiliies amugnst them.... still not cut down enough...

ok now we have another system that will rush through the possibilitys to stop all the purely / mostly static images from reaching production.. for a computer it can assume that any combination within a certain range, if similarly configured will result in a jumbled nothing of an image... this should cut out a multitude of useless images much faster than looking at them one by one.

it will be like looking through a single book (amungst a library of similar books) filled with jarbles of letters and maybe finding 14 pages mostly all text.. it might be nonesensical text but its text none the less.. and tearing out every page thats jarble. now if you know the progression of the systematic combinations of letters, you can figure out based on where the book begins how far into the book you'll have to look to find that jem of a page. essentialy you can automaticaly tear out the first 600 pages without having to even look at them becuase you already know that the progression of the combinations of letters doesn't reach anything good till the 601 page.

folow?

well a computer doing that would cut out enough to run down the possibilies that need to be produced.

but still might not be enough becuase there will be soo many duplicates that are the same shape but different colors ocuring at different stages of combinations. on set of images comes up in red, and then again in a lighter red, and again in blue and so on. i like robs idea of doing the shades of gray, although creating smaller images makes enlarging them nonesensical since the points of shade/color will also get enlarged (making a same size, but lower quality image) that and leaving the colorizing up to a human would be a horrible task to give to a person. it would be like having an extremely complexe paint by numbers, but without any numbers, and your range of possible colors is all of them. not the kind of effort one wants to go through to look at an image so i'll just stick with the grayscaling idea, but not all the way...

those sifting computers will only figure out the ranges of good images in greyscale combinations and send those ranges of good images to the next computer that actualy creates and colorizes the images.


that or we can illiminate color all together and just have the people deal with gray scale images. im fine with that.

oh and if anyone can simplify my explanation (assumin you understood it) with an equation, or any other way, feel free to post it up.

anyone know what the number of images would be if it were grayscaled?

[pat]

I assert that I already have such a database. If you specify your search algorithm, I will let it search my database.

Put another way, the actual database is superfluous if it is to contain every possible entry. The information content (entropy) is zero. Anything you search for will be there. Tell me which one you want and you've already got it.

[Keiji]

i like robs idea of doing the shades of gray, although creating smaller images makes enlarging them nonesensical since the points of shade/color will also get enlarged (making a same size, but lower quality image) that and leaving the colorizing up to a human would be a horrible task to give to a person. it would be like having an extremely complexe paint by numbers, but without any numbers, and your range of possible colors is all of them. not the kind of effort one wants to go through to look at an image so i'll just stick with the grayscaling idea, but not all the way...

hq9x is specially designed to keep the quality. It automatically rounds off corners, makes diagonal lines straight instead of pixelly, and gradients differences between similar colors.

As for the colorizing: The human doesn't color it in, they just mark out which parts are seperate from each other. The computer colors it in once this is done, in several combinations, and then the human picks which ones look good.

[batmanmg]

As for the colorizing: The human doesn't color it in, they just mark out which parts are seperate from each other. The computer colors it in once this is done, in several combinations, and then the human picks which ones look good.

hmmm... i think i'd prefer just looking at a greyscale. the concept is just as simple as the way they turned old black and white films into colors. i'd think that maybe a computer can do the job of diffining the separate parts of the picture itself.

and (since im a programming virgin) can anyone create a program that will do essentialy what i described on a much smaller scale? say a 8 x 8 array of black or white blocks, get diff numbers, letters and symbols like a calculator could. anything larger would be better for everyones understanding, and if you can work large enough that color would make an impact, by all means present its results. The some jumbled and some non jumbled.

^ ^ ^ if you'd like consider it a challenge, if not for personal achievement than for a better understanding of a simplisticly complicated senario.

I assert that I already have such a database. If you specify your search algorithm, I will let it search my database.

Put another way, the actual database is superfluous if it is to contain every possible entry. The information content (entropy) is zero. Anything you search for will be there. Tell me which one you want and you've already got it.

as for you pat. ummm what are you saying. that you already have a database of every image concevable by this setup, in a searchable format. or are you saying that theoreticaly any blank canvas is this database, and that the finished product is infact your search and result. the series of collors being the name of the peice, wich is also the peice itself.

if you mean you have a database, then how may i search through it. i can't simply tell you or even show you what i'm looking for. it would be like looking for a definition of a word online, and the only way to search for it, is to use the definition as your query. (sp?) so other than telling you what im looking for, is there any other way to find something?

well that is true, a blank canvase is in a sense, the same as this database. although searching it can be near impossible becuase it isn't pressented in an organized fashion. since there is no tangeble search feild we must assume that it lies in your mind. if you put every human being that exists, has existed, and in all likelyhood together in a room, and had them all create as many images they could in their lifetime. you wouldn't match in the least what the database im talking about would acheive. while i adore the human minds limitless potential, the fact is that the mind has very limited capability. mostly due to its limited time in existance. i could spend an entire lifetime creating but a single image that i desired to find. with a sophisticated search proceedure, or by random list searching, and a tangable field to search, one's search would be phenominaly more succesful. (i used search way too much, sry)

Edit: well both of those are rubbish on their own. not really achieving any gainful point. i took what pat said half the wron way both times. so heres what i have to say about it now.

if you are trying to remember the face of the man how held you up at gunpoint, would you be able to produce an image of him? usualy you would need a sketch artist, but even they can only create a depiction of him, a similar looking sketch, but if you put that image on the guys mirror, he'd hardly be fooled. and then theres the chance that you can't remember what he looked like, no matter how much you tried to think about it, you come up with bubkiss, an officer gives you a series of photographs, or a lineup to search through. and you may or may not be able to point him out. if you can't then you don't really know what your searching for becuase your starting out with a specific thing to be searching for. if you can remember him on sight, then it is that sight reconition that allowed your mind to conceive your prize. this shows that you need help to attain an actual rendition of what you want to find, as apposed to a knockoff copy.

[Keiji]

What he is saying is that what you search for is exactly what you get back. The database exists, but it is blank. If a database stores every possible combination for what it can store, there is no actual need to store anything at all.

[pat]

If a database stores every possible combination for what it can store, there is no actual need to store anything at all.

Unless those things are so time-consuming to come up with that one cannot generate them as needed. For example, all of the 30-digit primes could be a useful database. All of the 30-digit numbers could not.

[Keiji]

All the 30-digit primes would NOT be every possible combination for what it can store.

[batmanmg]

yes but you can consider the numbers that arent prime to be the jarbled images this database would exclude. the prime numbers the images we keep and search through.

[pat]

Okay, maybe I missed something. I thought the proposal was to *keep* every possible image in the database. You only want to keep non-jarbled images?

[Keiji]

No, the orginal argument was to store a database with all possible images. It was me that started the whole "no jarbled images" stuff.

[jinydu]

Except a painter doesn't use pixels, so you could have infinitely many paintings.

Painters do not intentionally use pixels, but there is a practical limit to how small a painter can paint something accurately. In any case, if all we care about is images that appear to be different to the human eye, this also sets a limit to the resolution of the pictures.

[batmanmg]

so does anyone think this is a feasable idea that some uber rich guy could put together and accomplish? or do you think that time won't permit?

to put a more mathamatical twist on it.

what would the limiting vairiables have to be to get the number down low enough for this to be accomplished?

whats the largest number of images we'll be able to creat at what speed.

how many images will have to be filtered out.

you can figure these out for grayscale images, or color, or both. preferably at 1024 x 768, but anywhere between that and 128x128 would work too.

[pat]

At this point, I still contend that the generating potential images is not the problem. Getting a rigorous definition of "non-jarbled" is the first problem (are any/all of the images below "non-jarbled"?). The second problem is how are you going to specify your search?



Do you expect to be able to search for "President George W. Bush" or "President George W. Bush having tea with three gila monsters"? If so, you've got your work cut out for you. Even if one had all possible images that everyone would agree fit the criteria, you'd still have bazillions of images (my rough guess would be more than there are stars in the Universe) returned in your search.

Easier to quantify would be the Infinite monkeys problem. You could pick a language like Lojban where it is easy for a computer to check whether the text is grammatically correct. And, you could make a database of every possible grammatical Lojban writing of N words or fewer (for your choice of N). Now, what? Why would you search this heap? Unless you've got some more stringent criteria than "all of the grammatically correct passages that contain the gismu 'basti' and 'ralju' within twenty words of each other" you're going to get back useless results.

If you were allergic to peanuts and I gave you seven tons of peanut M & M s with three non-peanut M & Ms mixed in, I've done you no favors.

[pat]

And, here's a "database" of all possible ("jarbled" or not) 256x256 binary images (with at least as much white as black). There is a Java Applet that lets you refine your search. The source code (and the "database") is in the JAR file.

Good luck finding this one:


It's in there.

[batmanmg]

overcomplicating the strictness of the searchers criteria is a probleme. almost none of those images where jarble, but you can easily define body structure and other objects structure with todays technology, the issue with creating what your looking for you limit the accuracy to your skills and knowledge of what you want to see.
you more than likely don't even know what your looking for ever until you've found it.
you know like i'll know what im looking for when i see it

but thats off the point, the point is that it will be difficult if not impossible to search through such a mess. you can't use word searches, mostly becuase your not searching for words, your searching for images, and if it was an analogy, then you wouldn't simply take two objects and search for them in the same picture, of course that would be obsurd.

but imagine you had writers block, and you had this idea for a sentence but you didn't know quite how to word it. or even if the words were quite the right ones. so you put down what you have and search for all the combinations / along with other words, within close proximity, from there you can continue to refine your search, by limiting it to gramaticaly correct statements (or if you want incoherent jarble, limit it to gramaticaly incorrect), then refine it more, you can refine the type of words that are added in adition to the searched words, (verb noun adjective adverb) , refine it some more, and so on and so on.

weither your search begins broad or unspecific, by refining your search, through shape, colors, hot colors vs cold colors orientation, solidarity, smoothness or texture, and soo many more, you can get the images themselves close to what you want, and limit the number of them.

the importance of this, would you, yourself, be able to creat either of those posted images? from scratch that is? or if those images, or in the firsts case, the original part of that image, never existed, would you be able to take parts of other images and creat that image? maybe something remotely like it, but hardly enough to be satisfactory. so how did you get those images, you had to search for them... the internet is simply this database, in an extremely limited form. your able to find images pretty good, but with a bigger selection you'd have better results, now with a bigger selection you'd need a better searching method, and super high tech searches can more than less likely accomplish that.

[Hugh]

This is an interesting concept to think about. Perhaps one could start out with the smallest grids and combinations and slowly work up to the higher ones. The thing to remember though, is that every possible image does have some possible meaning.

There is work being done with computers to identify letters in CAPTCHAs, and the steps are: (from the Wikipedia article at http://en.wikipedia.org/wiki/Captcha)
1. Removal of background clutter, for example with color filters and detection of thin lines.
2. Segmentation, i.e. splitting the image into segments containing a single letter.
3. Identifying the letter for each segment.

From the same page at Wikipedia: "Step 1 is typically very easy to do automatically. In 2005, it was shown that neural network algorithms have a lower error rate than humans in step 3. The only part where humans still outperform computers is step 2. If the background clutter consists of shapes similar to letter shapes, and the letters are connected by this clutter, the segmentation becomes nearly impossible with current software. Hence, an effective CAPTCHA should focus on step 2, the segmentation.

Neural networks have been used with great success to defeat CAPTCHAs as they generally are indifferent to both affine and non-linear transformations. As they learn by example rather than through explicit coding, with appropriate tools very limited technical knowledge is required to defeat more complex CAPTCHAs."

In the not too distant future, artificial neural networks will be much more capable than humans at all steps of identification of not only CAPTCHAs but other types of images as well.

What about looking at this idea from another angle? How about making a group of supercomputers capable of producing any image idea that you want from a detailed description?

You might say "show a car, now make it blue, put it on a hill of green grass, add a blue sky and a forest, etc. and make distinctions until you get what you want from the computer artist. It would be a great seller I'm sure...

[batmanmg]

for it to do that it would need a preset database of images much like the one already stated, just not all put together. and if your going to have all the angles of these objects, then your going to need really good 3d renditions that can be rotated and manipulated, wich are hard enough to make as it is, and a whole lot of them, but that would be more consuming, of time and space.

That and the computing power needed to put one of these images together would need to be imense, where as a search for an image is a much simpler process. You may not be able to find exactly what you want, but you'll be able to find something close enough you won't know the difference.

you can blast any of this off the page if its wrong.

But it is an extremely tempting idea. having a supercomputer be your own personal master painter.