[ExI] Unfrendly AI is a mistaken idea.

[Stathis Papaioannou]

On 05/06/07, A B <austriaaugust at yahoo.com> wrote:
>
> Stathis wrote:
>
> > "No we couldn't: we'd have to almost destroy the
> > whole Earth. A massive
> > meteorite might kill all the large flora and fauna,
> > but still leave some
> > micro-organisms alive. And there's always the
> > possibility that some disease
> > might wipe out most of humanity. We're actually less
> > capable at combating
> > bacterial infection today than we were several
> > decades ago, even though our
> > biotechnology is far more advanced. The bugs are
> > matching us and sometimes
> > beating us."
>
> Well, this is just splitting hairs growing on hairs
> but we will be in a good position to destroy all
> microorganisms and the useful earth within a couple
> decades, with something like molecular manufacturing.


Not if they develop resistance to the molecular manufacturing or whatever it
is; they do so to everything else, and they are doing so at an rate more
than matching our accelerating production of novel antibiotics, for example.


And it wouldn't require any genetic change to
> homosapiens. We must prevent that from happening of
> course, and we will. Microorganisms could never
> possibly destroy themselves as a species because they
> lack the intelligence to make it happen, unfortunately
> that's not the case with us.


Us destroying ourselves is not that different to other species' extinction
due to changes in the environment. We would be the ones changing our
environment but the same is the case when a species overpopulates and
consumes all its sources of food.

Do you honestly believe that the products of our human
> intelligence haven't conferred any survival or
> reproductive advantages, compared to other animals?


Obviously intelligence has, or it wouldn't have developed. But my argument
is that it appears to be just another trick that organisms can deploy, like
a better sense of smell or the ability to mutate quickly and develop
resistance to antibiotics.

> "I disagree with that: it's far easier to see how
> > intelligence could be both
> > incrementally increased (by increasing brain size,
> > for example) and
> > incrementally useful than something like the eye,
> > for example. Once nervous
> > tissue developed, there should have been a massive
> > intelligence arms race,
> > if intelligence is that useful."
>
> But the eye also evolved slowly. It likely began as a
> photo-sensitive skin pigmentation, that slowly evolved
> concavity, and so on. Human intelligence has only
> evolved once so far, because it was a much bigger,
> more complex, more unlikely "project".
>
> Below a certain threshold, I totally agree, a small
> incremental improvement in intelligence isn't likely
> to confer all that much benefit relative to the other
> animals. The likely threshold is the capacity to
> utilize tools (like sticks and rocks in multiple,
> varied ways) and to make tools. And I suspect that
> that one leap was *extremely* improbable, as evolution
> customarily never makes leaps but only baby-steps -
> and then only if they convey immediate  aggregate
> advantage. Imagine suddenly taking away from humans
> every invention we have ever made; would we really be
> much more "fit" than the other animals until we began
> making tools again? Probably not. Also, evolution
> could not have produced intelligence unless certain
> prerequisites were already in place. Magically giving
> a cactus human-level intelligence isn't likely to
> improve its survival or reproduction. The evolution of
> intelligence would require a means of perceiving the
> world (senses) and acting within in it (locomotion) -
> in such a way that the benefits of having more
> intelligence could be expressed in terms of advantages
> in survival or reproduction. And the parent animal
> would need to already have the physiology to allow the
> creation of tools: eg. standing semi-erect, and the
> infamous opposable thumb. That's why the cactus
> doesn't already have human-level intelligence, even
> though multicellular plants are way, way older than
> apes. So for these sorts of reasons, I consider the
> evolution of human intelligence as something of a
> miracle (in the strictly non-religious sense, of
> course). And something highly improbable, in all
> likelihood.
>
> > "It seems more likely to me that life is very
> > widespread, but intelligence is
> > an aberration."
>
> Yes, I meant that we are the first significant
> intelligence in this Universe, in my estimation.
> Intelligence is just an aberration like you say, but
> once it reaches human-level, it also happens to be
> extremely useful.
>

Either human-level intelligence is very difficult for evolution to pull off
or it isn't as adaptive as we humans like to think. You are arguing for its
difficulty; I still think a little bit of intelligence and a little bit of
tool manipulating wouldn't be that difficult, given the basic template of
mammals, birds, reptiles or even fish, and given predator-prey dynamics.

-- 
Stathis Papaioannou
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20070605/31c71885/attachment.html>