Zubon comments on Rationality Quotes September 2014 - Less Wrong

8 Post author: jaime2000 03 September 2014 09:36PM

You are viewing a comment permalink. View the original post to see all comments and the full post content.

Comments (379)

You are viewing a single comment's thread.

Comment author: Zubon 03 September 2014 10:43:35PM 12 points [-]

"You sound awfully sure of yourself, Waterhouse! I wonder if you can get me to feel that same level of confidence."

Waterhouse frowns at the coffee mug. "Well, it's all math," he says. "If the math works, why then you should be sure of yourself. That's the whole point of math."

-- Cryptonomicon by Neal Stephenson

Comment author: John_Maxwell_IV 07 September 2014 01:55:26AM 6 points [-]

This quote seems to me like it touches a common fallacy: that making "confident" probability estimates (close to 0 or 1) is the same as being a "confident" person. In fact, they're ontologically distinct.

Comment author: soreff 06 September 2014 04:28:51AM 4 points [-]

Was the context one where Waterhouse was proving a conditional, "if axioms A, B, C, then theorem Z", or one where where he was trying to establish Z as a truth about the world, and therefore also had the burden of showing that axioms A, B, C were supported by experimental evidence?

Comment author: VAuroch 22 September 2014 07:38:32AM 0 points [-]

Neither! The statement he is 'awfully sure of' is a probalistic conclusion he has derived from experimental evidence via Bayesian reasoning on the world's first programmable computer. Specifically, that statement is this:

"The Nipponese cryptosystem that we call Azure is the same thing as the German system that we call Pufferfish," he announces. "Both of them are also related somehow to another, newer cryptosystem I have dubbed Arethusa. All of these have something to do with gold. Probably gold mining operations of some sort. In the Philippines."

Part of the argument used to convince Comstock:

"But those places send out thousands of messages a day," Comstock protests. "What makes you think that you can pick out the messages that are a consequence of the Azure orders?"

"It's just a brute force statistics problem," Waterhouse says. "Suppose that Tokyo sent the Azure message to Rabaul on October 15th, 1943. Now, suppose I take all of the messages that were sent out from Rabaul on October 14th and I index them in various ways: what destinations they were transmitted to, how long they were, and, if we were able to decrypt them, what their subject matter was. Were they orders for troop movements? Supply shipments? Changes in tactics or procedures? Then, I take all of the messages that were sent out from Rabaul on October 16th--the day after the Azure message came in from Tokyo--and I run exactly the same statistical analysis on them."

Waterhouse steps back from the chalkboard and turns into a blinding fusillade of strobe lights. "You see, it is all about information flow. Information flows from Tokyo to Rabaul. We don't know what the information was. But it will, in some way, influence what Rabaul does afterwards. Rabaul is changed, irrevocably, by the arrival of that information, and by comparing Rabaul's observed behavior before and after that change, we can make inferences."