Re: Miracles, ETI, and rationality

Eliezer S. Yudkowsky (
Tue, 02 Nov 1999 03:54:13 -0600 wrote:
> It seems to me that accepting the premise of an interventionist alien
> civilization in residence on or near this planet, as several people
> have suggested, forces you to accept as a reasonable, plausible and
> significant probability that any number of miraculous, non-scientific
> events may in fact have been real, and may still be real.

It seems to me that you're having trouble accepting the proposition that anything is possible. In my philosophy, this is an engineering fact verging on an axiom; the fact that I cannot assign a 100% probability to any assertion trivially implies that I cannot assign a 0% probability to any assertion. I don't navigate qualitatively, ruling out branches of the future. I navigate quantitatively, using Occam's Razor.

Yes, anything is possible, but some explanations are more complex than others. Which is more likely; that the known forces of human memetics generated a legend, or that a Power intervened in this case, but not in others? The information contained in the memetics specification arises naturally from human psychology, evolution, and ultimately the laws of physics; on a fundamental level, it requires almost nothing that we haven't assumed already. Unless you can find an interventionist specification for some set of miracles that compactly explains why the Power intervened in these cases but not in other, similar cases, then your specification pretty much assumes what it's supposed to explain, requiring a much greater quantity of information to specify. Under the information-theoretic formulation of Occam's Razor, which is what I usually use, this would mean the first explanation is correct.

And while I agree that "Don't use Powers to explain things" has, historically, great value as a heuristic, cognition doesn't dictate external reality. If my model of the Universe predicts local and interventionist Powers, and a model that lacks said Powers is more complex and in fact contains characteristics that are obviously assumed solely to prevent said Powers from being there, then Occam's Razor - the rule underlying the no-Powers heuristic - takes over and makes the first explanation correct. Because the *real* heuristic is that you can't assume what you're trying to prove; that's why Power-based explanations don't *work*, they attribute the blueness of the sky to someone's intention that the sky be blue, and then refuse to explain the intention. That explanation contains as much information as what it explains. The simplest explanation is *not* "The lady down the block is a witch; she did it," because you haven't explained her motives. Reductive explanations are far more compact, informationally; you can explain complex surface behaviors by reference to simple elements and rules of interaction. And since we're in a reductive reality, and our world *is* made up of interacting elements, Occam's Razor works.

           Eliezer S. Yudkowsky
Running on BeOS           Typing in Dvorak          Programming with Patterns
Voting for Libertarians   Heading for Singularity   There Is A Better Way