If you can read this, either the style sheet didn't load or you have an older browser that doesn't support style sheets. Try clearing your browser cache and refreshing the page.

(Some Scientist)   Australian scientists invent true random numbers by splitting beam of light into two beams, measuring the power in each beam, then converting fluctuations into super-fast random number generator. Where's my flux capacitor?   (scientificcomputing.com) divider line 99
    More: Cool, random numbers, random number generators, virtual particles, flux capacitors, National University of Ireland, computer modeling, beam of light, quantum computer  
•       •       •

2386 clicks; posted to Geek » on 13 Apr 2012 at 6:03 PM (2 years ago)   |  Favorite    |   share:  Share on Twitter share via Email Share on Facebook   more»



99 Comments   (+0 »)
   
View Voting Results: Smartest and Funniest

Archived thread

First | « | 1 | 2 | » | Last | Show all
 
2012-04-13 08:27:35 PM

dittybopper: HR MSG NR 001 CK 24 0413 1646
BT
05231 05231 92092 73036 16767
09374 42290 96139 26217 64824
40156 03800 86873 83429 40353
19507 30392 00838 30914 23496
80422 61523 82287 16089
AR


Everything you need to read that is in my profile.


It would have been funny if you had replaced that last sentence with 'CSB'.
 
2012-04-13 08:39:21 PM
I still like the idea of a geiger-muller tube + source + comparator system simply because any geek can do it in their basement.

CS,B:

This probably isn't unique to the DOE, but it certainly fits them. Some of the sites I worked at had internal checkpoints that you had to pass through that triggered a random generator. If the buzzer went off when you went through... Whee! Happy Pat Down Time!

From what I understand they were of the system I described. Good luck fooling those.
 
2012-04-13 08:59:24 PM
Because light is quantised, the light intensity in each beam fluctuates about the mean. Those fluctuations, due ultimately to the quantum vacuum, can be converted into a source of random numbers. Every number is randomly generated in real time and cannot be predicted beforehand

This is an assumption based on an assumption. While, for immediate practical reasons, I see little reason this shouldn't be a safe thing to assume, ultimately, it is still a squared assumption.

Even assuming that their implementation is sound, the reality is that we're still assuming that what appears to be untestable pieces of quantum mechanical phenomena conform to the randomness that we currently assume to be true.

If science has taught us nothing else (which, of course, is nonsense), it has taught us not to assume that our assumptions are accurate. I have little more reason today to believe that our assumptions about quantum randomness are anymore correct than Plato did to assume that Aether was an accurate depiction of "the void".

God does not exist in gambling games anymore than he existed in the magical Aether.

/ also "quantisization" is not a thing unless __|---|__|---|__ and you're a poor speller
// sorry, I don't know how to do a square wave on Fark without an image....
 
2012-04-13 09:19:09 PM

Petit_Merdeux: 13 24 22557 3467464 9315

I'm doing this in my HEAD!

Is your mind BLOWN?

2976 69464 34646 48465 54646 4 6545451 897984346.54 54545 1484


Your brain hates 1, and loves 4, 5, and 6.

Humans are notoriously bad at picking random numbers. Most people don't know that most numbers start with "1".
 
2012-04-13 09:19:58 PM

impaler: Your brain hates 1 2, and loves 4, 5, and 6.


FTFM
 
2012-04-13 09:28:17 PM
The Science of Gamescience Dice:

Part One Link (new window)

Part Two Link (new window)
 
2012-04-13 09:31:29 PM
Can someone run this?

float fnum;
short *sptr;
int countArray[10];

memset(countArray, 0, sizeof(countArray));
sptr = (short *)&fnum;

for(int i = 0; i {
sptr[0] = rand();
sptr[1] = rand();
char str[32];
sprintf(str, "%f", fnum);
countArray[str[0]-'0']++;
}

for(int i = 0; i {
printf("amount that start with %d: %d\r\n", i, countArray);
}
 
2012-04-13 09:31:53 PM

ProfessorOhki: dittybopper: slayer199: Seems like a bit of overkill. Typical of mathematicians and physicists.

That being said, I can see this being handy for cryptography.

You know how I generate random numbers?

[i55.tinypic.com image 320x240]

Ten-sided dice. Cryptographically just as secure as any other method, and it only cost me a dollar a die (retail).

You know as well as I do, that unless those dice have perfectly uniform mass that they're not "truly random." They might be close enough random for practical purposes, but just engraving the digits on the faces means that they're not uniform.

/Unless the material was non-uniform in a way
//That specifically offset the differences in masses
///Nine, nine, nine, nine, nine, nine


It's not deterministic, in that you can't predict what the next number is. Plus, any imperfections in the individual dice are averaged by throwing all five at once to generate a five digit group.
 
2012-04-13 09:34:09 PM
or this rather...

float fnum;
short *sptr;
int countArray[10];

memset(countArray, 0, sizeof(countArray));
sptr = (short *)&fnum;

for(int i = 0; i < 2048; ++i)
{
sptr[0] = rand();
sptr[1] = rand();
char str[32];
sprintf(str, "%f", sptr);
countArray[str[0]-'0']++;
}

for(int i = 0; i < 10; ++i)
{
printf("amount that start with %d: %d\r\n", i, countArray[i]);
}
 
2012-04-13 09:35:54 PM

dittybopper: It's not deterministic, in that you can't predict what the next number is. Plus, any imperfections in the individual dice are averaged by throwing all five at once to generate a five digit group.


he did say they are "close enough random for practical purposes"
 
2012-04-13 09:42:42 PM
I'm a little disappointed that no one has taken a stab at deciphering the message I posted.
 
2012-04-13 09:44:32 PM

impaler: Another example:
Link (new window)


excellent link!
 
2012-04-13 09:51:59 PM

LazarusLong42: andrewagill: LazarusLong42: BKITU: Easier just to use the background microwave radiation to do it.

OK, crypto people, I've never seen a good answer for this: what's wrong with using microtime() (choose your system software) as a seed for a PRNG? Certainly the time at which you click the mouse is pretty noisy already if measured to the microsecond.

What if your system polls the mouse interface once every 300 microseconds, knowing that the user couldn't tell the difference even if the system could?

Well, true, obviously the system would have to poll that often, but clicking a button on a web page adds that noise to the system easily, and is no less secure than random.org


Anything that people do is full of patterns. Seriously. Which is one of the reasons why Dittybopper's dice are not random. The other reason is because they are not symmetrical.
 
2012-04-13 09:53:51 PM
Awesome! This will improve the size of my cock for sure!
 
2012-04-13 09:55:25 PM
25.media.tumblr.com

This should be enough to kick-start the Infinite Improbability Drive.
 
2012-04-13 09:55:30 PM

impaler: or this rather...

float fnum;
short *sptr;
int countArray[10];

memset(countArray, 0, sizeof(countArray));
sptr = (short *)&fnum;

for(int i = 0; i < 2048; ++i)
{
sptr[0] = rand();
sptr[1] = rand();
char str[32];
sprintf(str, "%f", sptr);
countArray[str[0]-'0']++;
}

for(int i = 0; i < 10; ++i)
{
printf("amount that start with %d: %d\r\n", i, countArray);
}

I ran it! farking awesome! Sorry it me so long to reply, I had to jack off afterwards.

 
2012-04-13 09:57:29 PM

dittybopper: I'm a little disappointed that no one has taken a stab at deciphering the message I posted.


I don't know how to convert it off the top of my head, but I recognize some of the control characters from morse code, yes?

If I hadn't been drinking all evening, I might try it, but since I have, I won't.
 
2012-04-13 10:04:27 PM

dittybopper: Plus, any imperfections in the individual dice are averaged by throwing all five at once to generate a five digit group.


I have to protest this one. If a number is more likely to have, say, three 2's than two 5's, then it is not random, regardless of where those digits lie in a 5 digit string. I'll wager that if you generated 300,000 of your numbers, pattens galore would present themselves (that would be a lot of work with dice, though). And granted, for small sample sets, they give you more random results than you would ever need.

/Sadly I have to run out the door so I can't continue participating in this discussion.
 
2012-04-13 10:04:42 PM

Splinshints: dittybopper: I'm a little disappointed that no one has taken a stab at deciphering the message I posted.

I don't know how to convert it off the top of my head, but I recognize some of the control characters from morse code, yes?

If I hadn't been drinking all evening, I might try it, but since I have, I won't.


Sort of. Here is the preamble:

HR MSG NR 001 CK 24 0413 1646 BT

Here is what it means:

"Here Message Number 001 Check 23 Date April 13 Time 4:46pm Break Text"
 
2012-04-13 10:11:54 PM
Random does not mean "has a uniform distribution." Dice can be random without being fair.
 
2012-04-13 10:15:07 PM
Remember when PEAR basically said they could prove people had telekenisis because of their random number generator?
 
2012-04-13 10:42:22 PM
dittybopper: It's not deterministic, in that you can't predict what the next number is.

But you could, given

1) a) You can calculate the physical effect of all the imperfections or b) You can create dice that are nearly perfect.

2) You can calculate the full physics of a toss (how bouncy is the surface, how hard are the dice etc etc).

3) You can accurately measure the forces acting on the die before a toss (make several observations on how the points move over time, and use that to extrapolate the vectors).

4) You don't have to calculate it in realtime (IE, before the dice hit stop moving),

IE, if someone rolled the dice, and only let the data gathered during point #3 above be the input to your computing, and the computer took an hour to produce an answer (long after the dice stopped rolling), would you still consider that a prediction (given that the computer only used data that was available during the roll to compute the result IE, the computer doesn't know what the result of the roll was save for its calculations so the problem wouldn't be with the predictions, but with the speed of said predictions.

The accuracy of the predictions depend on how accurate the pieces of your model are.

100% accuracy may never be achievable, but 70% or higher accuracy is enough to make us all millionaires in Vegas or on the stock market.

(of course, if we had prediction software that accurate, we could just become millionaires by selling it).
 
2012-04-13 10:52:09 PM

Hollie Maea: dittybopper: Plus, any imperfections in the individual dice are averaged by throwing all five at once to generate a five digit group.

I have to protest this one. If a number is more likely to have, say, three 2's than two 5's, then it is not random, regardless of where those digits lie in a 5 digit string. I'll wager that if you generated 300,000 of your numbers, pattens galore would present themselves (that would be a lot of work with dice, though). And granted, for small sample sets, they give you more random results than you would ever need.

/Sadly I have to run out the door so I can't continue participating in this discussion.


Let's say for example that die A is biased so that a 9 comes up 12% of the time instead of the expected 10%. This isn't useful cryptographically even if you could manage to recognize it. Plus, that particular die has only a 1 in 5 chance of being in position 1 for any given roll of 5 dice, so the *FAIR* odds of die A being in position 1 and rolling a 9 are 2%, and the biased odds of die A being in position 1 and rolling a 9 is 2.4%. That's a small enough difference that it would take an extremely large number of rolls to notice, and that would be submerged by the addition of plaintext to the key. Even if you had enough traffic to notice such a pattern, you couldn't exploit it cryptographically, because again, the dice are non-determinative. Just because die A is 2% more likely to roll a 9 than random chance would suggest, that doesn't mean that when you roll a 7, it has to come up a 9 the next roll.

Remember, I'm discussing this from a purely cryptographic standpoint. Manually generated one time pads created by rolling 10-sided dice are cryptographically secure.
 
2012-04-13 11:02:00 PM

lordargent: /determinist


Determinism suffers from a parsimony problem. To be a common determinist, you must believe that the universe is not parsimonious. A deterministic universe holds that any state of the universe can be derived from a) any other state and b) the time offset between your target state and your current state, and c) a function which accepts a state and a time offset as parameters, which is also a shorter-length bit-string than the universe.

But that raises the question: why do we experience events at all? A parsimonious universe wouldn't need to apply that function to generate events. The total bitlength of the function and a single state would obviously be less than a universe composed of n states for all n greater than one.

If the function were longer than the entire state of the universe, then you can still be a determinist- all future and past states of the universe can be computed, but not by an observer in our universe. Not only is this useless, but it's still not parsimonious- you've stated that the transition between two states in the universe is more complex than the universe itself.
 
2012-04-13 11:02:54 PM

NetOwl: Random does not mean "has a uniform distribution." Dice can be random without being fair.


True, but the strongest ciphers need a seed that is both random and fair. Dice rolls would be a relatively weak seed generator.

/seriously you people are arguing with someone who claims dice are the perfect RNG for any purpose? +10 trolling
 
2012-04-13 11:09:24 PM

andrewagill: The Copenhagen Interpretation holds that objective reality is the combined probabilities of all possible states.


Except for Oklahoma and Texas, obviously.
 
2012-04-13 11:15:07 PM

nekom: What advantage does a `TRUE' random number (if there even exists such a thing) have over a pseudorandom number in any application whatsoever?


TRUE randomness is impervious to pattern analysis. Pseudorandomness is only as sound as the ability to analyze the level to which it is not random. Which in some cases is pretty darn good. Today.

Link (new window)
 
2012-04-13 11:23:38 PM

dittybopper: Splinshints: dittybopper: I'm a little disappointed that no one has taken a stab at deciphering the message I posted.

I don't know how to convert it off the top of my head, but I recognize some of the control characters from morse code, yes?

If I hadn't been drinking all evening, I might try it, but since I have, I won't.

Sort of. Here is the preamble:

HR MSG NR 001 CK 24 0413 1646 BT

Here is what it means:

"Here Message Number 001 Check 23 Date April 13 Time 4:46pm Break Text"


Well, I was pretty close for a guy who spent three of the previous hours drinking
 
2012-04-13 11:31:25 PM

lordargent: Sure if you believe in that guy.


Er... I don't. I was just quoting Einstein.
 
2012-04-13 11:35:37 PM
Still no match for lead pipe crytpo.

www.vvsss.com
 
2012-04-14 12:26:58 AM
My random numbers are very old school, thank you.

www.danielsantiques.com
 
2012-04-14 12:41:43 AM

nekom: kingoomieiii: Well, the challenge is to get numbers that random electronically, and do so extremely fast.

What advantage does a `TRUE' random number (if there even exists such a thing) have over a pseudorandom number in any application whatsoever? Don't get me wrong, it's neat and all that but of what use is it?


Weather models and game design are a couple of things off of the top of my head. I know I had a programmer friend who was working o0n a better number randomizer for a while, and told me how it was a big deal.
 
2012-04-14 12:46:00 AM

lordargent: There's no such thing as 'random'.


I see you've never dealt with woman logic.
 
2012-04-14 12:54:03 AM

Propain_az: I ran it! farking awesome! Sorry it me so long to reply, I had to jack off afterwards.


could you post the output
 
2012-04-14 02:42:59 AM

aerojockey: NetOwl: Random does not mean "has a uniform distribution." Dice can be random without being fair.

True, but the strongest ciphers need a seed that is both random and fair. Dice rolls would be a relatively weak seed generator.

/seriously you people are arguing with someone who claims dice are the perfect RNG for any purpose? +10 trolling


There are, in my opinion, far more interesting applications of random processes than ciphers, but others disagree. I know quite a bit more about the other fields than I know about cryptography, so I tend to think of them first. (Of course, I think of this stuff as a big, messy application of measure theory, so I'm a little strange.)

Still, it's a common error that bugs me. A non-random die roll is not one where I use my "special" dice. It's when I roll behind the DM screen and lie about the result.
 
2012-04-14 03:07:30 AM
imgs.xkcd.com
 
2012-04-14 04:41:01 AM
OgreMagi : I see you've never dealt with woman logic.

It's too complicated for us to understand right now.
 
2012-04-14 06:18:02 AM
Huh. I didn't know Australia had technology that advanced.

/Unless maybe it was Space Australia

i40.tinypic.com

Go Space Broncos
 
2012-04-14 07:16:58 AM

wildcardjack: Still no match for lead pipe crytpo.

[www.vvsss.com image 448x274]


Actually, that's one of the great things about one time pads: They are resistant to what is more properly called "rubber hose cryptanalysis".

For example: Let's say you sent the message I sent, but using good one time pad practice, you burnt the paper pad and the worksheet you used to encrypt the message. You are then later arrested, and you are interrogated about what was in the message. If you tell them it was a personal message to an illicit lover, there is no way for them to prove you wrong. Hell, if you tell them it was a practice message of random numbers that don't make any sense, there is no way for them to prove you wrong. They can't make you give up the key, because you won't remember it: It's impossible. Hell, I've used that same key for a number of "one time pad" examples on Fark, going back at least a couple of years, and even if you beat me until nearly I died, I still couldn't remember it because it's a big string of random numbers.

Because there is no way for them to know if you are telling the truth or not, rubber hose cryptanalysis is impractical and a waste of time and resources. A clever person could give them a plausible, but innocent message the same approximate length or even a shorter one (to account for 'padding', a common practice with messages to hide stereotypical beginnings and endings). Hell, even a dullard could come up with a plausible summary of a bogus message, and there just is no way to prove them wrong without access to either the key, or the plaintext, both of which should have been destroyed immediately after the message was created.
 
2012-04-14 07:19:24 AM

OgreMagi: lordargent: There's no such thing as 'random'.

I see you've never dealt with woman logic.


Woman logic isn't random. It's alternately normal and perverse, according to a complex pattern.
 
2012-04-14 12:02:56 PM

NetOwl: There are, in my opinion, far more interesting applications of random processes than ciphers, but others disagree. I know quite a bit more about the other fields than I know about cryptography, so I tend to think of them first. (Of course, I think of this stuff as a big, messy application of measure theory, so I'm a little strange.)


Cryptography is the main use case for truly random and fair data; a good pseudo-random number generator with a seed gathered from entropy (e.g., dice roll :) is good enough for most random processes.

Most, I said. Some heavy problems in combinatorics might need better. For example, a PRNG that can generate a fair random shuffling of 10 thousand elements is beyond most existing PRNGs. But generally, crypto is the driving use case for true random data.
 
2012-04-14 12:28:41 PM
so its not a random number generator but simply a tool to measure mean density variations in atmospheric disturbances...

fail!
 
2012-04-14 12:34:08 PM

MessyDwarf: Nobody likes a smart-ass.


You never were invited to those sorts of parties, hm?
 
2012-04-14 03:45:04 PM

Petit_Merdeux: 13 24 22557 3467464 9315

I'm doing this in my HEAD!

Is your mind BLOWN?

2976 69464 34646 48465 54646 4 6545451 897984346.54 54545 1484


You said:
1 four times
2 four times
3 five times
4 twenty-one times
5 twelve times
6 twelve times
7 four times
8 four times
9 three times
0 zero times

You like 4.
 
2012-04-14 06:08:54 PM

poot_rootbeer: You like 4.


I like Scotch, too.
 
2012-04-14 09:24:38 PM

poot_rootbeer: You said:
1 four times
2 four times
3 five times
4 twenty-one times
5 twelve times
6 twelve times
7 four times
8 four times
9 three times
0 zero times

You like 4


God knows the internet teaches you to mouse with one hand and do everything else with the other.
 
2012-04-14 10:45:14 PM

lordargent: dittybopper: It's not deterministic, in that you can't predict what the next number is.

But you could, given

1) a) You can calculate the physical effect of all the imperfections or b) You can create dice that are nearly perfect.

2) You can calculate the full physics of a toss (how bouncy is the surface, how hard are the dice etc etc).

3) You can accurately measure the forces acting on the die before a toss (make several observations on how the points move over time, and use that to extrapolate the vectors).

4) You don't have to calculate it in realtime (IE, before the dice hit stop moving),

IE, if someone rolled the dice, and only let the data gathered during point #3 above be the input to your computing, and the computer took an hour to produce an answer (long after the dice stopped rolling), would you still consider that a prediction (given that the computer only used data that was available during the roll to compute the result IE, the computer doesn't know what the result of the roll was save for its calculations so the problem wouldn't be with the predictions, but with the speed of said predictions.

The accuracy of the predictions depend on how accurate the pieces of your model are.

100% accuracy may never be achievable, but 70% or higher accuracy is enough to make us all millionaires in Vegas or on the stock market.

(of course, if we had prediction software that accurate, we could just become millionaires by selling it).


You'd have to be watching the person rolling the dice to generate the pads, something that they aren't likely to let you do.
 
2012-04-14 10:54:37 PM

Hollie Maea: Anything that people do is full of patterns. Seriously. Which is one of the reasons why Dittybopper's dice are not random. The other reason is because they are not symmetrical.


Dice rolls having a pattern? Seriously?

This is so obviously wrong that I won't spend much time on it. If you think there is a pattern to my dice, find it in the pad in my profile.

It's true that at some level, they aren't symmetrical, and that over thousands, or tens of thousands of rolls, one particular number or numbers on any one of the die may come up very slightly more often than chance alone would dictate.

That doesn't help any potential cryptanalyst trying to break the messages, however. Some decidedly non-random one time pads have been used, but because they were non-deterministic, they still proved to be unbreakable.
 
2012-04-16 04:32:06 PM
I thought you math people thought there was no such thing as true randomness.
 
Displayed 49 of 99 comments

First | « | 1 | 2 | » | Last | Show all

View Voting Results: Smartest and Funniest


This thread is archived, and closed to new comments.

Continue Farking
Submit a Link »






Report