For “Harry Potter and the Methods of Rationality”, I’d like to make predictions that can be eventually verified publicly in time, but won’t tempt the author to change things to prevent it from coming true (as is the usual way of serial writers on the Internet). I’m therefore encrypting my prediction with the md5 hash function, so that afterwards it can be verified. (One difficulty is that I have the power to edit this comment; is there a way to make it obvious if I’ve done so, or store it in a less editable space?) Anyway, here goes:
July 16 (after Chapter 31): 28f9e3b2165344763c35514b473cb347
[03:35 AM] 215Mb$ echo "End of 32 only listed 3 options for Zabini. The fourth creates a 3-way tie. Zabini shoots himself as traitor." | sha1sum
09a3ee331d8900b7b7475b0a89911207672cbbda -
When the -n flag is added to the echo command (as suggested by ata) the hash matches.
So, unless orthonormal has exploited a security vulnerability on the Less Wrong server, orthonormal has expended a costly number of computing cycles to defeat the purpose of the hash function, I misunderstand the cryptographic guarantees provided by the hash function, or something equally unlikely, orthonormal really did write the following back on 25 July:
End of 32 only listed 3 options for Zabini. The fourth creates a 3-way tie. Zabini shoots himself as traitor.
I meant to say that orthonormal really did write the message back on the 25th with p=.99, and here are the three events that I am aware of that might cause my high level of confidence to be misplaced—the three largest sources of my remaining uncertainty, if you will.
If I assign p=.99 to an outcome and the outcome turns out to be false, it is a good bet that I misunderstood the question or misunderstood some proximate cause of the outcome, and in this case, the proximate cause I am most likely to misunderstand has to do with properties of hash functions.
You can have a third party create a cryptographically signed timestamp for you. For example, secure-timestamp.org will do this. This can only be falsified by getting the timestamping server’s private key or breaking its crypto algorithm. For things more important than Harry Potter predictions, you can have multiple third parties timestamp them for you, in which case falsification requires stealing all of their private keys.
MD5 is utterly utterly broken and recommended against for any purpose. Use SHA-1.
EDIT: I should mention that SHA-1 is also theoretically broken and may see a demonstrated break soon, but nothing like as problematic as MD5. Until SHA-3 is agreed, the SHA-2 functions are a good stopgap where you need better security.
Oh, good point. Fortunately, the actual importance of this whole endeavor is virtually nil, so it’s a good opportunity for me to learn how I ought to have done it better.
I think you’re confusing separate issues. The idea of having contests where two mathematicians would each issue challenge problems to the other was primarily an event in the Middle Ages. This idea then continued in some forms with people like Fermat (later Middle Ages or early Renaissance depending on how you define things) who would issue challenges to others to prove what they had already proven. What Orthonormal is doing actually more closely resembles a practice from the 19th century where when multiple people were racing to solve a problem first, they’d every so often deposit a sealed envelope with some reputable independent body to verify issues of priority.
You’re right, I was confused on the dates. I’m not referring to the sealed envelopes, but rather to the practice by which an alchemist/scientist/mathematician/etc would write a succinct summary of their discovery (i.e. “to transmute gold, add salt”) and write the letters in alphabetical order. They’d publish this somewhere public. Then, if someone else discovered the same thing, they would say “aha, but here’s my discovery” and point out how they came up with it ten years earlier by giving the “unhashed” version.
For “Harry Potter and the Methods of Rationality”, I’d like to make predictions that can be eventually verified publicly in time, but won’t tempt the author to change things to prevent it from coming true (as is the usual way of serial writers on the Internet). I’m therefore encrypting my prediction with the md5 hash function, so that afterwards it can be verified. (One difficulty is that I have the power to edit this comment; is there a way to make it obvious if I’ve done so, or store it in a less editable space?) Anyway, here goes:
July 16 (after Chapter 31): 28f9e3b2165344763c35514b473cb347
orthonormal predicts:
There, that’s less editable for you.
July 25 (after Chapter 32, prediction for Chapter 33):
SHA-1: f5721b3c6010973ee195dc160d0679477401a3df
Copying here to verify lack of editing in future.
Translation:
End of 32 only listed 3 options for Zabini. The fourth creates a 3-way tie. Zabini shoots himself as traitor.
Try echo -n.
Ah, you’re right. I forgot
echoappends a newline by default.When the -n flag is added to the echo command (as suggested by ata) the hash matches.
So, unless orthonormal has exploited a security vulnerability on the Less Wrong server, orthonormal has expended a costly number of computing cycles to defeat the purpose of the hash function, I misunderstand the cryptographic guarantees provided by the hash function, or something equally unlikely, orthonormal really did write the following back on 25 July:
Erm, yeah. I thought we all understood the hash scheme.
I see I have failed to communicate unambiguously.
I meant to say that orthonormal really did write the message back on the 25th with p=.99, and here are the three events that I am aware of that might cause my high level of confidence to be misplaced—the three largest sources of my remaining uncertainty, if you will.
If I assign p=.99 to an outcome and the outcome turns out to be false, it is a good bet that I misunderstood the question or misunderstood some proximate cause of the outcome, and in this case, the proximate cause I am most likely to misunderstand has to do with properties of hash functions.
Clear now?
If you edit the comment, an asterisk will appear after the time. Compare jimrandomh’s reply to the others.
Ah, clever.
EDIT: Let me see this for myself.
EDIT 2: Hey, it’s not working yet!
EDIT 3: Duh, I had to reload the page.
You can have a third party create a cryptographically signed timestamp for you. For example, secure-timestamp.org will do this. This can only be falsified by getting the timestamping server’s private key or breaking its crypto algorithm. For things more important than Harry Potter predictions, you can have multiple third parties timestamp them for you, in which case falsification requires stealing all of their private keys.
If you edit the comment a little asterisk will appear by the time stamp. Just make sure you don’t do that.
MD5 is utterly utterly broken and recommended against for any purpose. Use SHA-1.
EDIT: I should mention that SHA-1 is also theoretically broken and may see a demonstrated break soon, but nothing like as problematic as MD5. Until SHA-3 is agreed, the SHA-2 functions are a good stopgap where you need better security.
Thanks; I’ll do so in the future.
Did you include your own name in the text? If not, someone else can present the same hash and there’s no way to tell who came up with it.
Oh, good point. Fortunately, the actual importance of this whole endeavor is virtually nil, so it’s a good opportunity for me to learn how I ought to have done it better.
Can you say anything (without giving away the prediction) about when you would know if it is correct or not?
It’s my guess at (some features of) the ending.
You could use Prediction Book.
See, I’m even refusing to edit my badly written intro.
So, we’re returning to the days of math challenges? (i.e. the debate over the calculus?)
I think you’re confusing separate issues. The idea of having contests where two mathematicians would each issue challenge problems to the other was primarily an event in the Middle Ages. This idea then continued in some forms with people like Fermat (later Middle Ages or early Renaissance depending on how you define things) who would issue challenges to others to prove what they had already proven. What Orthonormal is doing actually more closely resembles a practice from the 19th century where when multiple people were racing to solve a problem first, they’d every so often deposit a sealed envelope with some reputable independent body to verify issues of priority.
You’re right, I was confused on the dates. I’m not referring to the sealed envelopes, but rather to the practice by which an alchemist/scientist/mathematician/etc would write a succinct summary of their discovery (i.e. “to transmute gold, add salt”) and write the letters in alphabetical order. They’d publish this somewhere public. Then, if someone else discovered the same thing, they would say “aha, but here’s my discovery” and point out how they came up with it ten years earlier by giving the “unhashed” version.
Hmm, you’re right, it would have been much more fun to do it that way.
Pastebin maybe?