[extropy-chat] A job for me?

[Adrian Tymes]

--- Eugen Leitl <eugen at leitl.org> wrote:
> On Mon, Nov 03, 2003 at 05:52:13PM -0800, Adrian
> Tymes wrote:
> > 
> > Such is the cycle of much that is AI at one point:
> 
> We don't have AI. AI doesn't mean artificial
> stupidity.

My point was that "AI" and "stupidity" are relative
terms.  What we have now is what would once have been
termed AI, even if we now state (with reason) it is
not.

> Isolated skills do not
> cumulate, nor do they magically integrate into a
> seamless whole -- so far.

There have been many other examples where individual
parts combine to something more than the sum of the
parts.  Synergies between different components, so
the overall whole is capable of tasks the components,
alone, are not.  There are indications that
intelligence as a whole might be the same - but we
have not even identified all the components as of yet.

> > First it is impossible.
> 
> It is.
> 
> > Then it is AI.
> 
> No, it's still impossible.

But fiction writers can at least write about
hypothetical solutions at this point, calling them
"AI", and their readers suspend disbelief instead of
claiming it can't possibly happen.

> People still enter text into
> editors when they want to
> solve a problem.

For certain problems, it is possible to solve without
touching a keyboard.  Granted, much of the solution
has to be presolved by people typing text into
editors, but this can be done far in advance for a
generic class of problem, then the particulars input
through other means.

> > When the ability to translate from rough natural
> > language descriptions to running code becomes free
> 
> We're still several years away from suffiently
> accurate speech recognition, a
> comparatively trivial task.

Did I say this was going to happen tomorrow?  ;P

> We don't have the
> hardware base to render anything but the flick part
> of that.

Yep.  We're a *long* ways away from the ultimate end
goal.  Not disagreeing with you there in the least.

> > and ubiquitous, we will have passed our current
> > Singularity - but perhaps be able to envision a
> new
> > one from there.
> 
> You don't see an event horizont when traversing a
> singularity. It's all in
> the eye of the external observer.

We're in violent agreement here.