[AR] Re: Testing Scaling Law
- From: Ed LeBouthillier <codemonky@xxxxxxxxxxxxx>
- To: arocket@xxxxxxxxxxxxx
- Date: Sat, 12 Nov 2016 07:09:51 -0700
On Fri, 2016-11-11 at 21:59 -0500, Norman Yarvin wrote:
The most useful ones, though, are actual *laws* -- laws of physics.
Scaling rules-of-thumb don't pack nearly the punch: they can be okay
as a sanity check, but since they're unreliable you can't use them as
a foundation to build on.
Sure, but models can be at different degrees of fidelity. As long as
you understand their premises, simpler models can be extremely useful.
Additionally, many things aren't actually *laws* in the physics sense
(except in, perhaps, the statistical sense). So how do you intuitively
understand things that aren't physical in nature? You use simplified
models that you can understand even though they don't represent *laws*
of physics.
And complexity makes the basic laws almost useless in understanding
aggregated effects. Sure, physics describes how a human works as much
as anything else. But it is useless to understanding human motivations
even though they, too, are the result of the same laws. You need to
abstract aggregrated effects in different ways. That's where abstract
rules that summarize things are useful even though they're not *laws*.
Other related posts:
