From: William Pearson (wil.pearson@gmail.com)
Date: Tue Oct 23 2007 - 20:41:36 MDT
On 24/10/2007, Eliezer S. Yudkowsky <sentience@pobox.com> wrote:
> William Pearson wrote:
> > I can't currently get around the problem that we haven't had any
> > instances of this happening.
>
> This is known as the "availability heuristic":
>
> http://www.overcomingbias.com/2007/09/availability.html
Untrue.
"The availability heuristic is judging the frequency or probability of
an event, by the ease with which examples of the event come to mind."
There are no examples of humans creating AI. That is all I am basing
my problem of finding a statistical argument for humans creating AI
on. I can readily see examples of the event coming to mind, but that
may not have any basis in its actual probability.
> You are trying to apply it in a way I've called the "absurdity heuristic":
>
> http://www.overcomingbias.com/2007/09/absurdity-heuri.html
Nope. I believe that humans creating AI is possible and probable, I
just want to find statistics that back up my belief. I'm not crying
absurd, just saying I don't see the statistical argument and evidence
for my belief.
> Also humans are a counterexample. If it seems to you that the brain
> is too magical to be duplicated by ordinary science, or that evolution
> is too powerful for human intelligence to match, that's a whole
> separate problem; but see also:
What it seems to me is neither here not there, I want cold hard
evidence for my belief.
Hmm, human capability vs evolutionary capability, that might be a
better thing to collect evidence of. The number of things we have
attempted to create that evolution has also created. This should be a
better predictor of what humans will invent, rather than the evidence
of the lack existence of the particular invention, which predicts new
inventions of that type will never be made.
However, it doesn't give evidence for believing humans will invent
things that haven't evolved such as lasers, microwave heating devices,
strong vacuum MNT or intelligences that alter their own code/hardware
and then recursively self-improve. So it is not the full answer to
being able to collect evidence for good hypotheses on what humans can
create or not.
It is really a shame there isn't (too my knowledge) a database of
inventors and their creations and failures, I could run some stats on
and look for common themes and percentage success rate, create
hypotheses and crunch some data.
Will Pearson
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:58 MDT