Not as sweet as a lava lamp, perhaps

December 9, 2008 4:34 AM

random.org offers true random numbers to anyone on the Internet. The randomness comes from atmospheric noise, which for many purposes is better than the pseudo-random number algorithms typically used in computer programs.
posted by Blazecock Pileon (44 comments total)

Hasn't the US government been using atmospheric noise as a source of "true" randomness?
posted by flippant at 4:44 AM on December 9, 2008

Additionally, Fermilab's HOTBITS .

HotBits are generated by timing successive pairs of radioactive decays detected by a Geiger-Müller tube interfaced to a computer. You order up your serving of HotBits by filling out a request form specifying how many random bytes you want and in which format you'd like them delivered. Your request is relayed to the HotBits server, which flashes the random bytes back to you over the Web.

posted by mikelieman at 4:48 AM on December 9, 2008

Is there any way to prove that these are "true random numbers"?
posted by beagle at 5:07 AM on December 9, 2008

Excellent, now I can get my random numbers delivered to me from a semi-trusted source over unencrypted http, instead of from that guy with the big scar in the alley behind Circuit City. It's a step in the right direction!
posted by Plutor at 5:08 AM on December 9, 2008 [7 favorites]

The items raised by the first three comments here are well covered by the OP's links (or elsewhere at random.org), especially beagle's. RTFA(s).
posted by intermod at 5:13 AM on December 9, 2008

Someone's bound to link to this xkcd comic eventually, so let's just get that out of the way.
posted by Johnny Assay at 5:17 AM on December 9, 2008 [4 favorites]

I don't believe this is a very good random number generator at all.
I asked for three random numbers between 1 and 10 and I got 1, 2 and 3 back.

Come on - that's a rubbish random number generator.
posted by seanyboy at 5:20 AM on December 9, 2008

True random? Surely the source for the numbers is a deterministic system of extraordinary complexity?
posted by vbfg at 5:22 AM on December 9, 2008 [1 favorite]

Is there any way to prove that these are "true random numbers"?

That's the problem with randomness. You can never be sure.
posted by dansdata at 5:24 AM on December 9, 2008 [5 favorites]

Will random.org be a random number generator site the next time I visit it?
posted by mandal at 5:25 AM on December 9, 2008 [2 favorites]

"The generation of random numbers is too important to leave up to chance."
posted by Vindaloo at 5:26 AM on December 9, 2008 [7 favorites]

That first link had other links which answered most of those.

e.g.: Secure Server HotBits Request

This page requests HotBits using the HTTPS protocol to encrypt the request for HotBits and the random data returned to you. This reduces the risk of your request being diverted to another site (because the request is encrypted with Fourmilab's certificate), and/or the random data being intercepted on its way to you.

But I thought no-one really needed to hammer the https server for demonstration purposes...

Other EXTREME links to satisfy your curiosity...

How HotBits Works
HotBits Hardware Description
HotBits Software Driver
Statistical Tests of HotBits Data

And it's not like John Walker is hiding under a rock or anything -- if you have any unresolved questions...
posted by mikelieman at 5:28 AM on December 9, 2008

A possible problem with random.org is the partial reliance on security through obscurity WRT the radio frequency. I imagine the secret frequency could be found with a little effort and some time using a computer-controlled transmitter, keeping the transmissions short and keep a close eye on the real-time statistics.
posted by exogenous at 5:32 AM on December 9, 2008

Is there any way to prove that these are "true random numbers"?

Define "random". If that definition eventually turns out to be "meets certain statistical criteria" then the answer to your question is yes.
posted by DU at 5:36 AM on December 9, 2008

I could listen to the sound generator all day.
posted by kuujjuarapik at 5:47 AM on December 9, 2008

I compare my random number generators by stacking column of numbers generated by the same method against other stacks generated by other methods. My god, some stacks are an inch taller!

Seriously, the notion of "chance" is a serious philosophical question, at least the last time I checked, which was in about 1982. British philosopher A.J. Ayer wrote a piece that appeared in Scientific American magazine in about 1965 entitled simply "Chance". I believe he came up with no less than six senses of the notion of chance.
posted by Tube at 6:01 AM on December 9, 2008 [4 favorites]

You know what's odd? For some reason, a word I know, but did not know the definition of, flashed in my head while reading this thread. I thought it occurred in the thread, maybe. No, it did not.

The word? Stochastic.

Weird.
posted by grubi at 7:21 AM on December 9, 2008

I want to see that, Tube. Annoyingly, there's like an 85-year gap in online coverage of Scientific American in the places I'm looking. It's 1965, volume 213 - Chance, by A. J. Ayer. A logician undertakes to sort out some of the several senses of this concept.

I want to read this so bad now. Can anybody dig it up?
posted by cashman at 7:23 AM on December 9, 2008

Can anybody dig it up?

?
posted by DU at 7:31 AM on December 9, 2008 [1 favorite]

Thanks.
posted by cashman at 7:44 AM on December 9, 2008

The site doesn't seem to indicate what the most random number between 1 and 10 is. I think it might be 2.
posted by Mr. President Dr. Steve Elvis America at 7:48 AM on December 9, 2008

Since these aren't encrypted, I'm worried that hackers are going to unrandomize my random numbers before I get them.
posted by lukemeister at 7:57 AM on December 9, 2008 [2 favorites]

Okay okay okay, I want to talk about the random sound generator. It just sounds like TV static, but it could sound exactly like Rick Astley. Right?
posted by roll truck roll at 8:02 AM on December 9, 2008 [4 favorites]

192540
posted by ALongDecember at 8:06 AM on December 9, 2008

I once took this astronomy class in high school. The teacher gave us tables of random numbers (from a book of random numbers) to plot as x-y coordinates on a sheet of paper. We were plotting our own star system. Each student got a different list of numbers. He graded them pass-fail.

So how did he grade them? If the stars were relatively evenly distributed around the paper, he knew the student cheated and it was a fail. I think the point was that even though we see patterns in the stars, that doesn't mean that they aren't randomly distributed across the sky (we lived too close to cities to be able to see the Milky Way).

In theory, a particular set of random numbers can be evenly distributed, but it is quite rare, and I'm sure he eliminated this possibility.
posted by eye of newt at 8:13 AM on December 9, 2008

I'm terrible at all things math, but just for fun I once decided to develop a random number generator as an intellectual exercise; my idea was to take several dozen sensors measuring a number of different things: the height in millimeters of the water in a river, the number of birds that flew past a camera, the sum of the harmonic values of the sound of tires on a bridge, etc. Arbitrarily multiply and divide them together, and then knock off all the whole numbers and everything else three to the right of the decimal point, and take whatever is left.

It's time consuming, but it's as close to truly random as I could come up with. As a plus, it would be cool to watch a machine collate all the data to come up with the number.
posted by quin at 9:09 AM on December 9, 2008

If anybody's in a hurry and doesn't have time to follow the link, here's a random number:

3
posted by signal at 9:20 AM on December 9, 2008 [3 favorites]

A possible problem with random.org is the partial reliance on security through obscurity WRT the radio frequency. I imagine the secret frequency could be found with a little effort and some time using a computer-controlled transmitter, keeping the transmissions short and keep a close eye on the real-time statistics.

I don't know what he's using, but if I was going to build such a setup, I'd tune it to 160.2 GHz, which is cosmic microwave background radiation. Essentially it's looking at slowly-cooling gas from the Big Bang, so what you get out of it is basically thermal noise. However, it's not totally uniform across the sky, so you'd have an additional variable that you could randomize (the direction you pointed the antenna) that would make any sort of prediction that much more difficult. Dial the gain on your receiver up high and you'd add some additional random analog noise.

It seems better than listing to an "empty" frequency, because there at least you would know the source of the signal you're getting, and know that it's random. It seems like it would be difficult to definitively prove, with equal certainty, the source of random-sounding noise on other bands.

I'm pretty sure that frequency range is basically blocked from being used for transmission (because it's so important to radioastronomy), so it would decrease the chance of something (passing airplane, guy next door on a cordless phone) accidentally interfering. That doesn't make it impossible that someone malicious could try to transmit a signal into it, but it would be difficult to transmit a signal that would overpower the CMB without causing immediate overloading of the receiver (especially with the gain jacked way up).

Unrelatedly: a while ago (probably 10+ years), I saw a dongle that you could plug into your computer's parallel port, that contained a true random number generator. I think it relied on thermal noise from a resistor. That would be a better bet for someone who didn't want to trust anything external for their random numbers.
posted by Kadin2048 at 9:22 AM on December 9, 2008

Hmmm.... I used to set GNU Backgammon up to get its dice rolls from Random.org. But the bot in that game always seemed to get exactly the rolls it needed to win. When it got three double-sixes in a row, gammoning me, I gave up on the game. Either the dice rolls weren't random, or the bot's "neural network" learned how to cheat.

(or I suck at the game and am just bitter)
posted by Thoughtcrime at 9:44 AM on December 9, 2008

Is there any way to prove that these are "true random numbers"?

If you mean random as in not having been caused by anything, then no, of course these aren't random. At the moment only certain quantum mechanical events appear to be random in this way - uncaused as it were, though this is a contentious matter. Obviously many things are practically "random" (in the sense of unpredictability) to us, because we can't (or haven't) worked out all the causal antecedents.

Information incompressibility is the most useful definition of randomness.
posted by phrontist at 9:47 AM on December 9, 2008

Since these aren't encrypted, I'm worried that hackers are going to unrandomize my random numbers before I get them.

Uhh...that could actually happen. Removing all the primes, say. Replacing the odds with evens.

But I don't think these random numbers are supposed to be omgsecure. Just...available online. So you don't have to code it up yourself in whatever language, you can just fetch a few.
posted by DU at 10:02 AM on December 9, 2008

In case someone needs it, here are some deterministic numbers.
posted by qvantamon at 10:24 AM on December 9, 2008 [2 favorites]

True random? Surely the source for the numbers is a deterministic system of extraordinary complexity?

Previously, Does God Play Dice?

Also previously:

In the end I think it is a philosophical debate more than a scientific one. Personally, I have come to think that events can actually be random, in the pure sense - not the unknowable but still somehow deterministic result of myriad other things happening in the universe, but actual capital-r random. Belief in pure randomness seems to correlate with post-modern thinkers and electrical engineers. Belief in unknowable determinism seems to correlate with modernist thinking and computer engineers.

posted by Chuckles at 10:40 AM on December 9, 2008

One attempt at an abstract definition was what Charles Bennett called logical depth: the number of computational steps needed to reproduce something from its shortest description.

If the Universe is seen as a computer, causally and irreversibly progressing towards greater entropy, then perhaps the numerical representation of cosmic radiation obtained via random.org is "truly random", in the sense that the Universe would need to be "rebooted" to repeat the same "computational steps", to reproduce the same digits in the same order. (In the same way, perhaps, that digits of π have to be computed to get from 3 to 3.1 to 3.14 to 3.141 to 3.1415, etc.)
posted by Blazecock Pileon at 11:05 AM on December 9, 2008

This is cooler than it has any right to be.
posted by fantabulous timewaster at 11:05 AM on December 9, 2008

Needs more eights.
posted by mandal at 11:08 AM on December 9, 2008

"Any one who considers arithmetical methods of producing random digits is, of course, in a state of sin. For, as has been pointed out several times, there is no such thing as a random number– there are only methods to produce random numbers, and a strict arithmetic procedure of course is not such a method." -John von Neumann
posted by bodywithoutorgans at 11:49 AM on December 9, 2008

I remember finding an old dictionary-size leather hardbound book of random numbers in the library.

It's true that many of the most common software based random number generators, when used to create two and three dimensional fields of graphic data, will show various dotted diagonal lines and such, that don't seem fully random.

There was a live video of a lava-lamp, which seeded a random process, available online at one time, as I recall.
posted by StickyCarpet at 11:55 AM on December 9, 2008

69 32 91 21 87 - self-explanatory I think you'll agree
posted by fallingbadgers at 12:14 PM on December 9, 2008

MetaFilter: As sweet as a lava lamp.
posted by not_on_display at 12:23 PM on December 9, 2008

Speaking of lava lamps...
posted by Pronoiac at 12:27 PM on December 9, 2008

About plotting generated pseudo-random numbers: You might be thinking of (Mefi's own) Dan Kaminsky's Phentropy, part of Paketto Keiretsu. The link to Phentropy leads to a dead end, but there's some video at the bottom of that page.
posted by Pronoiac at 12:36 PM on December 9, 2008

Here's another hardware source: radio noise as read by a wireless card.
posted by Zed_Lopez at 3:13 PM on December 9, 2008

Incidentally, LavaRand previously (at an old domain - I think they're the same project).

This is a double, by the way.
posted by Pronoiac at 3:56 PM on December 9, 2008