but then the claim of the philosophers responsible for working on this problem is simply incorrect. either that, or they have a despicably narrow-minded view of reason, in which case the thing they call “reason” is an uninteresting artifact of 21st century philosophical categorization. they might as well be using “reason n. dog poo,” for all the help this is in charting out the role cold calculation plays in this scenario, and my definition would at least require an interesting defense to justify the novel usage. this is a gamble, plain and simple, with fairly straightforward odds:
accuracy of algorithm: perfect
judgement of accuracy:
perfect: 1,000,000
imperfect: 1,000
accuracy of algorithm: imperfect and generous
judgement of accuracy:
perfect: 1,000,000
imperfect: 1,001,000
accuracy of algorithm: imperfect and miserly
judgement of accuracy:
perfect: 0
imperfect: 1,000
as there can be no question that the entire problem revolves around the nature of the predictor’s algorithm, decision theorists must be morons if this is an important problem to them. the only reason this may appear interesting is because in the way they framed everything, the whole thing defies their narrow definition of “reason”. this is a definition-oriented tug-of-war which dissolves as soon as you apply what eliezer says about splitting words. unfettered reason, being free to transcend the manner in which the problem has been set up, has no use for boundaries as artificial as conventional definitions. substitute “reason” with, say, “calculation”, and i no longer see what’s so special about it anymore.
but then the claim of the philosophers responsible for working on this problem is simply incorrect. either that, or they have a despicably narrow-minded view of reason, in which case the thing they call “reason” is an uninteresting artifact of 21st century philosophical categorization. they might as well be using “reason n. dog poo,” for all the help this is in charting out the role cold calculation plays in this scenario, and my definition would at least require an interesting defense to justify the novel usage. this is a gamble, plain and simple, with fairly straightforward odds:
accuracy of algorithm: perfect
judgement of accuracy: perfect: 1,000,000 imperfect: 1,000
accuracy of algorithm: imperfect and generous
judgement of accuracy: perfect: 1,000,000 imperfect: 1,001,000
accuracy of algorithm: imperfect and miserly
judgement of accuracy: perfect: 0 imperfect: 1,000
as there can be no question that the entire problem revolves around the nature of the predictor’s algorithm, decision theorists must be morons if this is an important problem to them. the only reason this may appear interesting is because in the way they framed everything, the whole thing defies their narrow definition of “reason”. this is a definition-oriented tug-of-war which dissolves as soon as you apply what eliezer says about splitting words. unfettered reason, being free to transcend the manner in which the problem has been set up, has no use for boundaries as artificial as conventional definitions. substitute “reason” with, say, “calculation”, and i no longer see what’s so special about it anymore.