Rationality, Time and Minds

From: Eliezer S. Yudkowsky (sentience@pobox.com)
Date: Sat Sep 22 2001 - 14:13:36 MDT


*Bzzzt.*

Not the whole thread, but some aspects of it.

It is possible to discuss religion in the light of SL4 topics. But it has
to be done carefully. Just discussing something ultrapowerful and
ultraintelligent, even omnipotent and omniscient, is not the same as a
license to switch on the faith-activated areas of the brain and say
anything that enters your head.

The reason I am invoking the great and awesome power of *bzzt* in defense
of this principle is not just that this kind of fuzzy thinking is like a
cheese grater on my nerves, but because it can easily kill a list. I am
not banning religious speculation - but I am banning low-quality religious
speculation. (Speculation which contains incorrect
spelling/grammar/punctuation will automatically be deemed to be low
quality.)

I'm sorry to have to do this so early into the discussion, but no traffic
is better than bad traffic.

Those of you who are long veterans of the war against fuzzy thinking
already know what I mean by all this. Those list members who don't come
from scientific backgrounds - I expect we'll have at least a few - may be
somewhat confused as to what the fuss is about, or may think I'm saying
that all religious discussion is automatically unscientific. Well, to
some extent, it is, in the sense that if you have a really logically
rigorous discussion about religion all the religious ideas will end up
being shot down, because the religious ideas are in fact wrong and we have
enough evidence to know it. But that doesn't mean that it isn't possible
to discuss religion logically - or that people who disagree with my
evaluation of religion can't post logical, rigorous opposing viewpoints.
(Samantha Atkins can still post.)

So what exactly the problem? The influence of anticipation on perception,
as I would put it; in plain English, "wishful thinking". Bad enough when
people engage in flagrant rationalization in support of a conclusion they
would wish to be true; even worse is just posting the conclusion without
bothering with the rationalization.

Reasoning is a tool for telling us things we didn't previously know. Not
for producing sentences that make us feel warm and fuzzy inside. It is
acceptable to post a warm and fuzzy sentence, but it has to be supported,
and the standard to which that support is held will not be lowered.

The Universe runs on its own rails; so must rationality. If, in the due
course of time, unbiased reason happens to naturally produce a warm and
fuzzy conclusion, then by all means post the conclusion and the
reasoning. The idea that pain and death can be eliminated is an example
of a conclusion that has been produced by both rationalization (in
religious prophecy) and unforced reasoning (we have the technology); the
former pseudologic does not invalidate the independent real logic. (It
does mean that normative observers will legitimately be more skeptical of
the person professing such a heartwarming conclusion, until it has been
established to their satisfaction that the logic really is independent of
the desired answer.)

A subtle example of rationalization might be asking the question: "Why is
there pain and suffering if it doesn't really need to be that way?", and
then offering the answer "Because otherwise life would be boring." To
meet the standards of the SL4 mailing list, the question must be phrased
as follows: "Postulating that our world is currently under the control of
an ultrapowerful entity, and supposing that this entity has values that
have at least some intersection with our own, why do we observe pain and
suffering?" Now, the simple answer is that our world is NOT under the
control of an ultrapowerful entity. If under the control of an
ultrapowerful benevolent entity, we don't expect to see this world; we see
this world, and so the "ultrapowerful benevolent entity" hypothesis is
disconfirmed. If you go on to postulate a possible exotic set of rules
under which a benevolent superintelligence might create/allow pain and
suffering, you have to treat this as an "additional assumption" under
Occam's Razor and note that the probability of the final scenario is
reduced accordingly, not say "And so we see that pain and suffering are
really OK, and we're all still cradled in God's arms."

Many philosophers throughout history, faced with a world of pain and death
from which there was apparently no escape, or having grown up with the
commanded assumption that there exists a superintelligence running our
world, have rationalized the existence of pain and suffering. This is to
be expected. But there is no reason why such theories should be allowed
to contaminate SL4. I don't say that the *conclusions* are forever
forbidden, because there might be *independent* logic supporting them.
But the theories as theories are known to involve twisted logic, and that
the pseudologic is not acceptable here.

-- -- -- -- --
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 - 04:00:37 MDT