Library of babel

35 replies [Last post]
tomlukeywood
Offline
Joined: 12/05/2014

i recently discovered the Library of Babel (source code will be released later this year according to author)
which uses a algorithm to generate every possible 3200 character combination that can ever exist
and you can index exact combinations with a number

and it also dose the same thing for images

so this is unbelievably cool as effectively everything ever said within 3200 characters can be generated by the algorithm
or any possible image of everything that did or did not happen

but the best part of this i was thinking is if you have this algorithm then all you need is the correct number and you can generate any possible string of text or image or any data

so is this not some kind of ultimate compression? why is this not in use already?

also dose this not bypass copyright law as you can just distribute a number and then everyone can generate themselves that data

so for example if you give this number to the library of babel algorithm it will make a image of nintendo owned mario:
http://pasted.co/538d4974

GNUser
Offline
Joined: 07/17/2013

That looks really cool in a... useless kinda way :-P
Thanks for sharing!
We could use it to make cool logos :-P

GNUser
Offline
Joined: 07/17/2013

Could it be used to generate some kind of onion services names? Maybe could be integrated to get more "reasonable" names... Dunno much about that.

quantumgravity
Offline
Joined: 04/22/2013

It is mathimatically impossible to create information out of nothing;
if you give your friend one number, that's it, he just has this one number.
Of course inside the program there can be an index, so if you give it the number "12" it will look up "text number 12" in some internal database.

I also don't really get what's so new or cool about it.
Generating any possible character combination was always possible with pc and only limited by computional cost. This certainly hasn't changed, maybe the algorithm makes it more efficient or something.

tomlukeywood
Offline
Joined: 12/05/2014

"It is mathimatically impossible to create information out of nothing"
it dose not create information out of nothing the algorithm effectively creates ALL information within a limit
but dose not have to create it all at once but by ensuring that when a certain number is entered into the algorithm the same data is always produced you can store the program to make that data just as a number

it would be good to see the source code when the author release it though to see how they do it

GNUser
Offline
Joined: 07/17/2013

Sorry but, basically what I think they do is create every string that begins with "a" like "aaa", "aab", "aac"... "abc" and so on. They move on to another letter. What counts is the amount of computing power they used.

But how to turn that string you posted into an image??

tomlukeywood
Offline
Joined: 12/05/2014

you need to enable JS to use it but its this page:
https://babelia.libraryofbabel.info/slideshow.html

you just click on the number at the bottom and change it

onpon4
Offline
Joined: 05/30/2012

I don't know about this thing being used to replace compression. The number you gave as an example is almost 1 MB of text. I don't know what Mario image that is, because I don't understand how to see it, but for example, the largest PNG image I find in ReTux at a glance is less than 500 KB, and that's an 800x600 painted background image. So it seems likely that the text is larger than the actual image.

Magic Banana

I am a member!

Offline
Joined: 07/24/2010

Precisely. The number certainly is longer that the text or the image compressed with a proper algorithm (say GZIP for texts, PNG for images).

onpon4
Offline
Joined: 05/30/2012

As for copyright... if you legitimately just randomly find an image in this thing that happens to be identical to a copyrighted image, then sure, that would let you use the image. But the important word here is if. That's a pretty bold requirement. If you actually already knew of the image and specifically searched for it, which is the only way I can imagine ever getting the image you want, I don't think the courts would treat you well.

tomlukeywood
Offline
Joined: 12/05/2014

but if someone else found it and then distributed it across the internet would that number effectively be illegal?

onpon4
Offline
Joined: 05/30/2012

The problem is the chance of that happening is astronomically small. An individual could very easily spend every second of his life looking through the collection of random images and not find a single useful one. And what about making images fit together? Forget making animations this way!

To wit, in an 16-bit bitmap image of the size 64x64 (not terribly large, medium color depth, and with no alpha channel), each pixel can be one of 65,536 colors. 65536*64*64 = 268,435,456 possible images. Assuming you could check one image every second and never take any breaks, it would take you about 8.5 years to look through all of them.

It's possible to use randomness to generate things that are useful, but you have to employ rules to only return results that are likely to be useful. See Sfxr, for example, which -- rather than generating random bits -- randomizes various settings for how to generate a sound effect. I honestly can't think of a way to do this for images in any useful manner.

tomlukeywood
Offline
Joined: 12/05/2014

no i mean someone found the number using a premade copyrighted image and then distributed the number would that number then be illegal?

onpon4
Offline
Joined: 05/30/2012

That's hard to say. I think it would be pretty much the same deal as distributing a torrent file; you're not actually distributing the work in question, but you're distributing information on how to obtain it.

In any case, I don't think courts would accept that as an excuse for redistributing the image found there.

Magic Banana

I am a member!

Offline
Joined: 07/24/2010

It is the same information in another format. The copyright still applies. Otherwise, it would be like saying that the copyright on a text does not apply once you compress the text (e.g., using GZIP).

JadedCtrl
Offline
Joined: 08/11/2014

Every comment on this thread is already in the Library of Babel.
You might as well stop typing-- no matter what you write, or how hard you try, a machine has already beaten you to it.

GNUser
Offline
Joined: 07/17/2013

Wrong... we have to keep typing so we write more than 3200 characters, and that way we will beat the machine! ;-P

DIEEEEEE TERMINATOOOOOR!!! (sorry, geeky moment happening)

JadedCtrl
Offline
Joined: 08/11/2014

If you type more than 3200 characters, than the rest of what you type is simply on a different page.
But we have a few things that we do better than these damn robots- capital letters, apostrophes, hyphens, etcetera!
... One day, it'll be able to do those characters too, though. :(

GNUser
Offline
Joined: 07/17/2013

wait, can't the machine also make all possible combinations with those?

Magic Banana

I am a member!

Offline
Joined: 07/24/2010

It can. Any character can be represented as a sequence of 0 and 1. See 'man ascii' for instance.

In fact, any information can be represented as a sequence of 0 and 1. It is even the way it is physically stored on a disk.

JadedCtrl
Offline
Joined: 08/11/2014

I mean, it has the potential to (and the epic goal of the Library is to utilize every ASCII character) but the algorithm only supports 28 characters right now.

GNUser
Offline
Joined: 07/17/2013

If only the algorithm could also throw away the "nonsense" images, and retain only images that actually look like "something". I am sure Google image search software could be used to do it.

Actually doing the same with text search would be interesting ;)

Magic Banana

I am a member!

Offline
Joined: 07/24/2010

As far as I know, there is no known algorithm that would recognize a "sensible" picture from random pixels.

GNUser
Offline
Joined: 07/17/2013

There is software to recognize faces for example (all cameras use it) and Google has done a lot of work in image recognition. I believe it could at least be used to take a lot of "nonsense" images out of the pool we would have to search afterwards.

tomlukeywood
Offline
Joined: 12/05/2014

that would be weird!

JadedCtrl
Offline
Joined: 08/11/2014
GNUser
Offline
Joined: 07/17/2013

Hum... ok, are those for real? Or is it just a well played prank? How did you find those??

tomlukeywood
Offline
Joined: 12/05/2014

there is a search feature in the library if babel

quantumgravity
Offline
Joined: 04/22/2013

Yeah but wait a sec..
you can just search for "GNU manifesto" and you will get hundred thousands of results, don't you?
I mean, the program created all possible character combinations.
So among the millions of pages containing "nonsense...nonsense...gnu manifesto...." somebody has to pick the real gnu manifesto.
How is this done?

JadedCtrl
Offline
Joined: 08/11/2014

I copied and pasted the first 3200 characters of the manifesto into the search.

quantumgravity
Offline
Joined: 04/22/2013

Yeah but see, you had to feed it with the whole text in order to get it back.

My point is:
What is talking?
It's picking a combination of letters from the very large amount of possibilities in order to express yourself.
Just because this huge amount of possibilities is now written down somewhere doesn't change anything.

GNUser
Offline
Joined: 07/17/2013

True but if you think about it... there must be some "pearls" waiting to be discovered. I mean... imagine that I would now make an emotional speech about something, something deep philosophical and profound... it is already written there, and if we could "stumble" on it we would have found some amazing text that could motivate people without it ever being written by "anyone".

Like I said, it's interesting in a "useless kinda way".

Magic Banana

I am a member!

Offline
Joined: 07/24/2010

The time to generate a pearl is prohibitive. See https://en.wikipedia.org/wiki/Infinite_monkey_theorem and https://en.wikipedia.org/wiki/Infinite_monkey_theorem#Random_document_generation in particular for practical experiments.

andermetalsh
Offline
Joined: 01/04/2013

Just play a little with Perl and Hailo (cpan, use cpan2deb to build the deb package) .

GNUser
Offline
Joined: 07/17/2013

I remember a donald duck comic book where they found a library that contained all the books ever written and they wanted to find locations of gold and eventually discovered that the library had all the truths and lies ever written, so they had so much information that it was like having no information at all.
Kinda similar to this...