From: Dan Fabulich (daniel.fabulich@yale.edu)
Date: Mon Mar 13 2000 - 01:54:55 MST
'What is your name?' 'Eliezer S. Yudkowsky.' 'Do you deny having written
the following?':
> Dan Fabulich wrote:
> >
> > So AI and space travel are totally compatible goals. d.Otter may have
> > good reasons to not want you to build the seed at all, but that's a
> > totally unrealistic thing to hope for. So where's the argument left?
>
> I'm totally on board with both. In fact, I've explicitly proposed doing
> both. What worries me is the part where Otter wants to outlaw AI.
Outlaw AI?
Notice that Eliezer's plan to Singularity starts with a better programming
language and a really good debugger and works up from there. Are you
gonna outlaw debuggers? Just "smart" debuggers? (What the heck good is a
DUMB debugger?) Or maybe self-optimizing compilers? After all, a really
good compiler might lead to dancing.
Note as well, once the computers get big enough, you can just run a brute
force search for an AI.
A koan: Consider a lossy search algorithm like a "genetic" algorithm,
searching a large space for an optimal solution while neglecting some
search space. What's the best lossy search algorithm for isolating the
best lossy search algorithm?
> You know and I know that as long as humanity survives, the creation of
> AI is inevitable; that doesn't mean that the creation of AI can't be
> delayed until nanocatastrophe either wipes out most of Earth's
> population (which I'd view as a partial loss) or all life in the Solar
> System (a total loss).
Sure, it's inevitable EVENTUALLY, though he might, at least in principle,
upload first, if he sets enough roadblocks (like outlawing AI, etc.)
Well, this is no time for advocacy, then... the race is on!
Of course, as it just so happens, he's got a much harder task set for him,
with or without AI laws. (I can claim this without argument; we'll just
see who's right, eh?)
May the best idea win!
-Dan
-unless you love someone-
-nothing else makes any sense-
e.e. cummings
This archive was generated by hypermail 2.1.5 : Fri Nov 01 2002 - 15:27:20 MST