From: Eliezer Yudkowsky (sentience@pobox.com)
Date: Sun Oct 24 2004 - 17:04:10 MDT
Jeff Medina wrote:
> Eliezer said: "I have little use for conclusions that are absolutely
> certain given their assumptions. I want conclusions that are
> correct."
>
> You speak to this effect frequently, as though there were some way
> around assumptions and uncertainty. Would you mind elaborating on this
> source of objective, infallible truth to which you demand all
> suggestions to live up?
Sigh. I mean the following: I don't care if you prove that, based on some
assumption or other, you can fly when you step off a cliff, if you actually
plummet to the ground and go splat. My goal is to outguess the mysterious
agent "reality" which produces the actual experimental results we compare
against our predictions.
I'm a Bayesian, Medina. I don't assign probabilities of 0 or 1 to
anything. My math goes kerplooey when it tries to handle infinite
certainties. The suggestion that I ever claimed anyone could be absolutely
certain of anything is base calumny verging on deliberate insult.
> I'm baffled as to how you can make statements
> like the above quoted one, when there is no means of doing better than
> "conclusions that are absolutely certain given their assumptions". All
> reasoning requires two sorts of assumptions; (1)
> axioms/premises/empirical data (the place from which reasoning starts)
> and (2) rules of inference (methodology that is *assumed* to be valid
> for producing new truths from old ones).
Some great mathematician, I forget who, said: "Insofar as our ideas are
about reality, they are not certain; insofar as they are certain, they are
not about reality." Whatever is true of all possible worlds can never
provide any information about *which* possible world we live in.
I want conclusions that are strongly probable given previous observations,
and which, in fact, turn out not to send me shrieking off a cliff to my
doom. This latter property I call "correctness" and you can substitute
"not too grossly wrong" if you like.
This matter of certainty given assumptions seems to me no commendation of a
theory, if its conclusion fails against fact.
-- Eliezer S. Yudkowsky http://intelligence.org/ Research Fellow, Singularity Institute for Artificial Intelligence
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:49 MDT