so who is there left to trust?
February 14, 2021 2:27 PM Subscribe
Basilisk collection - From Wikipedia, the free encyclopedia is a work of unfiction by Blackle Mori.
According to the author, unfiction is "fiction that insists it's real."
According to the author, unfiction is "fiction that insists it's real."
The author had a process running for over a year to generate the hashes used in the story. Respect.
posted by phooky at 2:51 PM on February 14, 2021 [8 favorites]
posted by phooky at 2:51 PM on February 14, 2021 [8 favorites]
Sierra Hotel echo (space) Oscar November Lima Yankee (space) whisky alpha November tango echo delta (space) alpha (space) papa uniform bravo Lima India Sierra Hotel India November golf (space) delta echo alpha Lima (space) brave uniform tango (space) tango Hotel India Sierra (space) bravo echo Charlie alpha Michael Echo (space) foxtrot alpha November tango alpha Sierra Yankee (semi-colon) (space) uniform November foxtrot India Charlie tango India Oscar November (stop)
posted by parmanparman at 2:53 PM on February 14, 2021 [4 favorites]
posted by parmanparman at 2:53 PM on February 14, 2021 [4 favorites]
One popular explanation for the creation of the Basilisk collection is the "happy family" theoryI thought Tolstoy said it was happy families that were all alike, and the unhappy families which were identifiably unique and deterministic hashes?
posted by Fiasco da Gama at 2:55 PM on February 14, 2021 [18 favorites]
The use of 'basilisk' is a shame, since the same name was used for an earlier series of excellent SF stories, several of them pseudo-real, by David Langford.
posted by Hogshead at 3:25 PM on February 14, 2021 [7 favorites]
posted by Hogshead at 3:25 PM on February 14, 2021 [7 favorites]
I think it benefits the piece to reference Langford's Basilisk and also Roko's Basilisk.
posted by save alive nothing that breatheth at 3:36 PM on February 14, 2021 [10 favorites]
posted by save alive nothing that breatheth at 3:36 PM on February 14, 2021 [10 favorites]
I suspect the use here of 'basilisk' is meant to evoke Roko's basilisk rather than Langford basilisks. That is, to spoil the subtext, that the hashes might not be produced by human beings at all.
posted by Pyry at 3:37 PM on February 14, 2021 [4 favorites]
posted by Pyry at 3:37 PM on February 14, 2021 [4 favorites]
I understand about 1/2 of this story and love what I can grasp.
Reminds me of that JG Ballard story which is an index for a book that doesn't exist.
posted by doctornemo at 4:02 PM on February 14, 2021 [2 favorites]
Reminds me of that JG Ballard story which is an index for a book that doesn't exist.
posted by doctornemo at 4:02 PM on February 14, 2021 [2 favorites]
I don't understand this bit:
posted by Joe in Australia at 4:39 PM on February 14, 2021 [1 favorite]
In general, for a hash of N bits, a difficulty value of 2^K would require on average 2^(N-K) hash evaluations to find a corresponding nonce value.That implies you need fewer hash evaluations as the difficulty increases? But then the author says:
Using the language of proof-of-work systems: the inversions use the challenge string "basilisk:N:" where N is a 10 digit number and the difficulty value is no less than 2^168.Assuming the challenge string is ASCII text, it's 8*18=144 bits long. I'm not sure it makes sense to have a hash longer than the challenge string, but even so, at some (high) difficulty you'd have 2^(N-K)=1. That is, it would take a single hash evaluation to find a partial collision. I'm not a cryptographer, but is there a typo or am I misunderstanding things?
posted by Joe in Australia at 4:39 PM on February 14, 2021 [1 favorite]
This completely went over my head but I love the structure of it, the care put into the fake sources is especially nice. Something about this kind of "found archival footage" type content really scratches an itch.
posted by Emily's Fist at 5:11 PM on February 14, 2021
posted by Emily's Fist at 5:11 PM on February 14, 2021
For the "difficulty value" described in the story, lower numbers are more difficult, as the goal is to find a hash that's less than this value.
posted by save alive nothing that breatheth at 5:27 PM on February 14, 2021 [2 favorites]
posted by save alive nothing that breatheth at 5:27 PM on February 14, 2021 [2 favorites]
I loved this story -- thanks for the post!
That implies you need fewer hash evaluations as the difficulty increases?
Right, in this context a lower difficulty number is harder, because you're trying to get a hash value that is less than the difficulty target. The easiest difficulty number is 2^256 (all hash outputs are less than this), and the hardest difficulty is 0 (no hash outputs are less than this). I think it's done this way just because it's prettier than using greater-than, because successful hashes will start with a bunch of zeroes.
Then later I think you're right that "the difficulty value is no less than 2^168" is a little confusing -- it must mean no less as in "no easier", not no less as in "no lower."
Assuming the challenge string is ASCII text, it's 8*18=144 bits long.
The length of that part of the string doesn't really matter. To explain what's going on a little more, the first hash the entity in the story provided is this:
basilisk:0000000000:ds26ovbJzDwkVWia1tINLJZ2WXEHBvItMZRxHmYhlQd0spuvPXb6cYFJorDKkqlA 0000000000000000000000161b9f84a187cc21b172bf68b3cb3b78684d8e9f17
Meaning they solved this computational challenge: given the string "basilisk:0000000000", discover 64 alphanumeric symbols you can append to that string, such that
Doing this once is, as far as we know, very expensive; doing it 500,000 times is effectively impossible. (It's also a lovely subtle way to walk up to that sense of wonder I'm always hoping for from scifi -- wait OK but what would that mean?)
posted by john hadron collider at 5:38 PM on February 14, 2021 [5 favorites]
That implies you need fewer hash evaluations as the difficulty increases?
Right, in this context a lower difficulty number is harder, because you're trying to get a hash value that is less than the difficulty target. The easiest difficulty number is 2^256 (all hash outputs are less than this), and the hardest difficulty is 0 (no hash outputs are less than this). I think it's done this way just because it's prettier than using greater-than, because successful hashes will start with a bunch of zeroes.
Then later I think you're right that "the difficulty value is no less than 2^168" is a little confusing -- it must mean no less as in "no easier", not no less as in "no lower."
Assuming the challenge string is ASCII text, it's 8*18=144 bits long.
The length of that part of the string doesn't really matter. To explain what's going on a little more, the first hash the entity in the story provided is this:
basilisk:0000000000:ds26ovbJzDwkVWia1tINLJZ2WXEHBvItMZRxHmYhlQd0spuvPXb6cYFJorDKkqlA 0000000000000000000000161b9f84a187cc21b172bf68b3cb3b78684d8e9f17
Meaning they solved this computational challenge: given the string "basilisk:0000000000", discover 64 alphanumeric symbols you can append to that string, such that
sha256(sha256("basilisk:0000000000:<your symbols>")) is less than 2^168.Doing this once is, as far as we know, very expensive; doing it 500,000 times is effectively impossible. (It's also a lovely subtle way to walk up to that sense of wonder I'm always hoping for from scifi -- wait OK but what would that mean?)
posted by john hadron collider at 5:38 PM on February 14, 2021 [5 favorites]
The only things missing (arguably) are functioning edit histories and talk pages.
posted by nosewings at 5:41 PM on February 14, 2021 [4 favorites]
posted by nosewings at 5:41 PM on February 14, 2021 [4 favorites]
... oh but hopefully what it means is A Deepness in the Sky style aliens or a superintelligent computer sending an inscrutable message, right? It could just be a mathematician deciding that shorting bitcoin was the most money they could make with their sha256 break, and in real life I guess that would be the most likely explanation, but it would be a bit of a letdown.
posted by john hadron collider at 5:48 PM on February 14, 2021 [1 favorite]
posted by john hadron collider at 5:48 PM on February 14, 2021 [1 favorite]
Reminds me of some of Scott Aaranson's writing on big numbers.
how many values of the [Busy Beaver] function is it possible for us to know? Where exactly is the precipice at which this function “departs the realm of mortals and enters the realm of God”
posted by rhamphorhynchus at 6:37 PM on February 14, 2021 [1 favorite]
how many values of the [Busy Beaver] function is it possible for us to know? Where exactly is the precipice at which this function “departs the realm of mortals and enters the realm of God”
posted by rhamphorhynchus at 6:37 PM on February 14, 2021 [1 favorite]
https://cybre.space/@SuricrasiaOnline/105732606867125186 is a response from the author about the difficulty value part
Rest of the site's worth a look too, I enjoyed https://suricrasia.online/library/
posted by xiw at 7:36 PM on February 14, 2021 [4 favorites]
Rest of the site's worth a look too, I enjoyed https://suricrasia.online/library/
posted by xiw at 7:36 PM on February 14, 2021 [4 favorites]
Love this!!
posted by lock robster at 10:12 PM on February 14, 2021
posted by lock robster at 10:12 PM on February 14, 2021
I love this. Very well done.
I have one thing I can't figure out, though. The fact that the author really did find those nonces suggests something is fudged in the story, no?
If 5 of the 500 million can actually (in our and the author's reality) be found in one year on a single PC, then surely all 500 million are not beyond the reach of "all computers created by humanity ever" in a similar time scale. But the difficulty of finding one nonce is stated as 2^(n-k), or 2^(256-168)=2^88 hash calculations in this case. 2^88 is roughly 3*10^26 hash calculations, which if done one year would be a rate of ~10^19 per second -- far, far beyond any single computer's capacity (which fits the story, of course). So while one could get lucky and find a nonce sooner, it's exceedingly unlikely to get that lucky... nevermind five times in a year.
Either the difficulty is overstated in the story (but I don't see how it is; the math checks out) or the author didn't really find those nonces with that difficulty level (but they all check out too!). So I'm missing something. Help? [If the answer is "The story's author is actually a superhuman intelligence" that's... cool, I guess. Hi.]
[keeps checking things before posting...] ...Oooohhhhhhhh. I see. I love that, too! Um, spoilers ahead, kinda? Like maybe if you either want to retain your amazement or if you want to figure it out yourself? Either one of those might be more fun than reading the rest of this comment.
The hashes don't check out, but in a clever, tricky way! If you run the verification command in the story (with the
Now, I like to save time when I manually compare hashes by not comparing every character, but rather just the first several and the last several characters, which are easier and faster to eyeball. After all, the probability of a hash matching the start and the end but not the middle is pretty darn small! Well, there are different values of "small" when we're dealing with hash probabilities.
So the author... oh, the tricksy author here... They doctored their hashes. See, the hashes of the story's example strings do match the hashes given in the story in their first many non-zero characters, and they do match the last ten or so characters. Someone like me who bothers to check that they're legit -- and who like me is used to not checking every character of a hash -- will be fooled. But the real hashes have extra characters in between those start and end chunks, and crucially the real hashes also start with many fewer zeros. The real hashes are the same as the story hashes but with extra characters added, balanced out with fewer zeros to start. The hashes in the story start with 22 zeros. The actual hashes that the example strings produce start with 11 zeros. That means that, effectively, the real hashes have a massively easier difficulty level. An achievable-in-reality difficulty level.
So the author used a year's worth of computation just to find values they could use to fool the few readers who actually try to check the hashes. Just so those few readers might buy into the story that little bit more. That's nice.
posted by whatnotever at 10:32 PM on February 14, 2021 [16 favorites]
I have one thing I can't figure out, though. The fact that the author really did find those nonces suggests something is fudged in the story, no?
If 5 of the 500 million can actually (in our and the author's reality) be found in one year on a single PC, then surely all 500 million are not beyond the reach of "all computers created by humanity ever" in a similar time scale. But the difficulty of finding one nonce is stated as 2^(n-k), or 2^(256-168)=2^88 hash calculations in this case. 2^88 is roughly 3*10^26 hash calculations, which if done one year would be a rate of ~10^19 per second -- far, far beyond any single computer's capacity (which fits the story, of course). So while one could get lucky and find a nonce sooner, it's exceedingly unlikely to get that lucky... nevermind five times in a year.
Either the difficulty is overstated in the story (but I don't see how it is; the math checks out) or the author didn't really find those nonces with that difficulty level (but they all check out too!). So I'm missing something. Help? [If the answer is "The story's author is actually a superhuman intelligence" that's... cool, I guess. Hi.]
[keeps checking things before posting...] ...Oooohhhhhhhh. I see. I love that, too! Um, spoilers ahead, kinda? Like maybe if you either want to retain your amazement or if you want to figure it out yourself? Either one of those might be more fun than reading the rest of this comment.
The hashes don't check out, but in a clever, tricky way! If you run the verification command in the story (with the
openssl commands), it produces a hash that looks just like the one in the story. On a quick glance: "yup, looks legit." Repeat the command for the other four example strings, and they check out too.Now, I like to save time when I manually compare hashes by not comparing every character, but rather just the first several and the last several characters, which are easier and faster to eyeball. After all, the probability of a hash matching the start and the end but not the middle is pretty darn small! Well, there are different values of "small" when we're dealing with hash probabilities.
So the author... oh, the tricksy author here... They doctored their hashes. See, the hashes of the story's example strings do match the hashes given in the story in their first many non-zero characters, and they do match the last ten or so characters. Someone like me who bothers to check that they're legit -- and who like me is used to not checking every character of a hash -- will be fooled. But the real hashes have extra characters in between those start and end chunks, and crucially the real hashes also start with many fewer zeros. The real hashes are the same as the story hashes but with extra characters added, balanced out with fewer zeros to start. The hashes in the story start with 22 zeros. The actual hashes that the example strings produce start with 11 zeros. That means that, effectively, the real hashes have a massively easier difficulty level. An achievable-in-reality difficulty level.
So the author used a year's worth of computation just to find values they could use to fool the few readers who actually try to check the hashes. Just so those few readers might buy into the story that little bit more. That's nice.
posted by whatnotever at 10:32 PM on February 14, 2021 [16 favorites]
I'm genuinely shocked at my inability to tell 22 zeros from 11 zeros by eye.
posted by Omission at 1:09 AM on February 15, 2021
posted by Omission at 1:09 AM on February 15, 2021
I'm genuinely shocked at my inability to tell 22 zeros from 11 zeros by eye.
posted by Omission
Name checks out.
posted by Joe in Australia at 2:54 AM on February 15, 2021 [4 favorites]
posted by Omission
Name checks out.
posted by Joe in Australia at 2:54 AM on February 15, 2021 [4 favorites]
... oh but hopefully what it means is A Deepness in the Sky style aliens or a superintelligent computer sending an inscrutable message, right? It could just be a mathematician deciding that shorting bitcoin was the most money they could make with their sha256 break
I'm just going to chalk this up to another attempt by Yay Newfriend to get the attention of their bestest friend Roko Basilisk. Or maybe it was something Melon did by accident.
posted by RonButNotStupid at 5:39 AM on February 15, 2021 [1 favorite]
I'm just going to chalk this up to another attempt by Yay Newfriend to get the attention of their bestest friend Roko Basilisk. Or maybe it was something Melon did by accident.
posted by RonButNotStupid at 5:39 AM on February 15, 2021 [1 favorite]
I agree with xiw. The rest of that site is also fun.
posted by doctornemo at 7:25 AM on February 15, 2021
posted by doctornemo at 7:25 AM on February 15, 2021
« Older Beloved Palestinian Poet and Memoirist Mourid... | Ice, ice baby Newer »
This thread has been archived and is closed to new comments
posted by seanmpuckett at 2:44 PM on February 14, 2021 [3 favorites]