Re:Bayes's Theorem: What's the Big Deal?
by loup-vaillant on Thursday January 07, 2016 @08:56AM
(#10857893)
You want me to read it? Okay.
I don't like the introduction, mostly because I believe what he ridicules to a degree (AI is an existential risk, I have enjoyed HPMOR more than Rowling's work, and I'm not quite sure that making lots of money and give it away is worse than working directly for whatever cause we want to support).
Bayesian grace
Bayes T-shirt means certain asshole. Well at least it made me smile.
I agree that Bayes' theorem is not that notable. The product rule, more fundamental, is more representative of the laws of probability (the actual basis for bayesianism).
Of course, the "formula for the perfect brain" is computationally intractable…
The association between Bayesianism and Neo-Liberalism looks like an ad-hominem attack —doesn't apply to me, at least.
I can't comment on this "Bayesian revolution". But I'm already suspicious of the whole essay at this point, and cannot trust this paragraph.
Amazing Bayes
What the author fails to acknowledge here is that to the extent it can apply, Probability theory is that amazing. There are, like proofs of it working very broadly, and it relies on very few axioms. (Jayne's work in Probability Theory: the Logic of Science leaves little doubt about that.)
Absence of evidence is evidence of absence. Often very weak evidence, but evidence nonetheless. Whoever believe otherwise doesn't understand probability theory. (Maybe the author conflated "evidence" and "proof", or "evidence" and "strong evidence"?)
The correct application of probability theory is computationally intractable in many cases. I can see how it would be unworkable for historians, who have to juggle with many many kinds of evidence. Not having read Richard Carrier however, I'm not sure this objection is not yet another strawman.
Accusing Bayesianism to be responsible for confirmation bias is ridiculous however. Confirmation bias does not follow probability theory. It often follows an incorrect application of it however.
Less Wrong
Okay, I'm out. The insults are too blunt, came too soon, without any justification so far. SIAI (now MIRI) as a doomsday cult is a strawman. As for LessWrong, Eliezer specifically warned about the dangers of using rationalist's tools for rhetoric purposes —how your own biases can increase when you know about biases.
I'm not taking the effort required to search for and debunk any justification that might come later. He just used up my patience.
I don't like the introduction, mostly because I believe what he ridicules to a degree (AI is an existential risk, I have enjoyed HPMOR more than Rowling's work, and I'm not quite sure that making lots of money and give it away is worse than working directly for whatever cause we want to support).
Bayesian grace
Bayes T-shirt means certain asshole. Well at least it made me smile.
I agree that Bayes' theorem is not that notable. The product rule, more fundamental, is more representative of the laws of probability (the actual basis for bayesianism).
Of course, the "formula for the perfect brain" is computationally intractable…
The association between Bayesianism and Neo-Liberalism looks like an ad-hominem attack —doesn't apply to me, at least.
I can't comment on this "Bayesian revolution". But I'm already suspicious of the whole essay at this point, and cannot trust this paragraph.
Amazing Bayes
What the author fails to acknowledge here is that to the extent it can apply, Probability theory is that amazing. There are, like proofs of it working very broadly, and it relies on very few axioms. (Jayne's work in Probability Theory: the Logic of Science leaves little doubt about that.)
Absence of evidence is evidence of absence. Often very weak evidence, but evidence nonetheless. Whoever believe otherwise doesn't understand probability theory. (Maybe the author conflated "evidence" and "proof", or "evidence" and "strong evidence"?)
The correct application of probability theory is computationally intractable in many cases. I can see how it would be unworkable for historians, who have to juggle with many many kinds of evidence. Not having read Richard Carrier however, I'm not sure this objection is not yet another strawman.
Accusing Bayesianism to be responsible for confirmation bias is ridiculous however. Confirmation bias does not follow probability theory. It often follows an incorrect application of it however.
Less Wrong
Okay, I'm out. The insults are too blunt, came too soon, without any justification so far. SIAI (now MIRI) as a doomsday cult is a strawman. As for LessWrong, Eliezer specifically warned about the dangers of using rationalist's tools for rhetoric purposes —how your own biases can increase when you know about biases.
I'm not taking the effort required to search for and debunk any justification that might come later. He just used up my patience.