Re: AI:This is how we do it

From: Anders Sandberg (asa@nada.kth.se)
Date: Mon Feb 18 2002 - 06:18:32 MST


On Mon, Feb 18, 2002 at 07:16:49AM -0500, Eliezer S. Yudkowsky wrote:
> Eugene Leitl wrote:
> >
> > On Sun, 17 Feb 2002, Eliezer S. Yudkowsky wrote:
> >
> > > People are asking the same questions they were asking in 1999. I could
> > > understand if there'd been progress and yet not progress toward Friendly AI,
> > > but this is just stasis. Why?
> >
> > You realize of course that Friendly AI is your baby. Most people outside
> > this circle haven't heard about it, and I personally find Friendliness a
> > mythical property. <insert the usual arguments here>
>
> I know. But I would have expected progress somewhere, and I would have
> expected that people would find it useful to, e.g., distinguish between
> supergoals and subgoals, regardless of beliefs about outcomes, or for that
> matter, regardless of beliefs about the nature of interaction between
> supergoals and subgoals. I'd expected that at least the *concepts* would be
> useful.

First, are we talking about people on this list or people discussing AI
in general? If the first, then I would agree we should take into account
concepts that has been extensively debated on this and related forums.
But if we look at the whole field, we must be aware that our ideas have
not been spread much and hence will not affect it. This is why I think
it is essential for anybody who really wants the world to take their
ideas into account to write publications that are published in the
academic mainstream or at least somewhere where they will be read - no
matter how brilliant one's ideas are, if they are forever relegated to a
website hidden under a self-invented terminology they will not have much
effect on other people's thinking. It goes not just for your AI ideas,
but for transhumanism in general.

In many AI architectures I don't know if its is meaningful or easy to
speak about supergoals and subgoals, for example neural network based
reinforcement learning agents where goals emerge and compete
dynamically.

-- 
-----------------------------------------------------------------------
Anders Sandberg                                      Towards Ascension!
asa@nada.kth.se                            http://www.nada.kth.se/~asa/
GCS/M/S/O d++ -p+ c++++ !l u+ e++ m++ s+/+ n--- h+/* f+ g+ w++ t+ r+ !y


This archive was generated by hypermail 2.1.5 : Fri Nov 01 2002 - 13:37:39 MST