From: Eliezer S. Yudkowsky (sentience@pobox.com)
Date: Tue Apr 20 1999 - 01:03:15 MDT
Lyle Burkhead wrote:
>
> I wrote:
> >
> > 1. What is fantastic is the image of goo spreading over the planet,
> > everything in its path. Microscopic life won't have an advantage over
> > large-scale life in the future any more than it does now. Hard life, like
> > natural life, will come in all sizes, from microbes to Godzilla, and all
> > forms, from algae to insects to mammals, plus other forms that have not
> > been thought of. The ecosystem of the future will be at least as complex
> > and diverse as the ecosystem of today. The world isn't going to be any
> > more gooey in the future than it is now.
>
> To which Eliezer Yudkowsky replied, in one of the most amazing
> non sequiturs to appear on the list in a long time -- at least in the last
> couple of hours --
The Non Sequitur Society: We don't make sense, but we do like pizza.
> > I'm so glad to know that submarines will never outswim fish.
>
> What on earth is this in reference to?
>
> Hard life will, as I have said repeatedly, have an advantage over soft
> life. However, microscopic hard life won't have an advantage over
> large-scale hard life any more than bacteria have an advantage over
> mammals. Hard life will come in all sizes, from microbes to submarines.
> An ecosystem of hard organisms won't be any more gooey than an ecosystem of
> natural organisms.
Please understand, Lyle, that I don't give a damn whether or not an
ecosystem of hard organisms is gooey. What I care about is the human
race being prematurely wiped off the face of the Earth. It doesn't
matter whether the replicators can "really" compete with a rabbit or if
this is just an illusion of scale; if they can outcompete the cells that
make up the rabbit, the rabbit ceases to exist.
As for the non-sequitur, I assumed you were arguing that goo, for some
reason, couldn't eat us, because it wouldn't be any better a competitor
than the cells. I answered this with the old analogy about "Can a
submarine swim?" If a bullet can kill us, why not goo? And if goo can
kill us, but only in a fashion which makes it analogous to a biological
ecology, who cares? We're still dead.
> > Yes; submarines might be more of a problem if the population
> > was not kept down by the submarine-eating denizens of the deep.
>
> If there were predatory submarines, then there would indeed be
> submarine-eating denizens of the deep, i.e. other submarines.
But there aren't. Again, I assumed that you were talking about a reason
not to be afraid of goo. Certainly, by calling the goo scenario absurd,
you're doing a good job of creating the impression that we can all relax
and stop worrying. If you really mean that we ought to call it the
"microscopic a-life creatures scenario", but we're still all going to
die, then I really must question whether your efforts are promoting the
proper attitude.
> They would
> eat each other. They would also be subject to attack by smaller a-life
> creatures. Microscopic a-life creatures ("goo") would be more interested
> in attacking submarines than in attacking us. In other words, the submarine
> population would be kept in check just like the population of any other
> organism is kept in check.
Yes, but in point of fact, the submarine population is *not* kept in
check, and there's no reason to suppose that grey goo will be either.
> I wrote
> > 3. Finally, the other fantastic thing about the grey goo scenario is that
> > it is all supposed to happen overnight.
>
> Eliezer interrupted at this point and said
> > Says you. It could just as well happen after a decade.
> > Straw man. Nobody specified overnight.
>
> The word "overnight" is a figure of speech, meaning "suddenly." My
> understanding of the goo scenario is that even if it takes a decade to
> start, once it does start it is supposed to happen very quickly indeed.
Sure. Imagine bacteria that doesn't stop when it reaches the edge of
the petri dish. If you really think it will take a minimum amount of
time, that's a technical question. Prove it. Besides, let's suppose it
is slow. I don't understand why it's better to run from a creeping tide
of black death, hacking off fingers infected by dust specks along the
way, until you drop from exhaustion and are agonizingly consumed.
> The goo is going to "eat our facility and condemn us all to horrible,
> agonizing deaths!" That scenario is fantastic, and the word "overnight" is
> not out of place in describing it. Nor is the word "overnight" out of
> place in describing a scenario in which an entire new kind of life is
> designed and built in a decade.
No, *you* want an entire new kind of life. *You're* the one who's been
saying that mere self-replication is life by definition, and now you
want to get us to buy that it will have to be as complex as biological
life. Technical question. Prove it.
And remember, long ago there arose a single molecule that could
self-replicate. By *accident*. That's why we're all here.
As for making fun of a scenario where everything changes in a decade, I
very much regret that technological progress is not obliged to conform
to your desired pace. Where was the Web a decade ago? Technical
question. Prove it.
> Here I am trying to get a handle on question #1 above, by considering the
> difficulty I have had in redesigning my own cells. Ignoring this, Eliezer
> chose to focus on the word "overnight." So who's attacking a straw man?
> Instead of taking the word "overnight" literally, why don't you reply to
> the statement "Replicators are complex, and that complexity isn't going to
> go away"?
It doesn't have to go away. Complex problems have been solved. Drexler
thinks they'll probably have an assembler by 2013. Think it will take
longer? Explain why I should believe you instead of Drexler. Technical
question. Prove it.
And no, *I* don't have to prove it. Eric Drexler and a lot of other
people I respect tell me to worry, so I worry. You don't want me to
worry, because you don't think it can happen so fast. Why? Because
you're making fun of it. Your analogies can just as well prove that I
don't have this amazingly complex computer on my desk, or that the Web
didn't happen in a decade.
If calibration works so well, let's see you calibrate the stock market.
In fact, can you give me even *one* example of calibration working?
*Ever*? *Any* event in history that could have been "calibrated" by an
event as distant as the ones you show here? Your "calibration" would
assume that it takes as long to go from Neanderthal to Homo sapiens as
from amoeba to Neanderthal, or that it would take as long to go from
email to Web browsers as from the printing press to telephones.
Or let me put this as simply as possible: I don't care what you decide
to "calibrate" against. Eric Drexler wrote _Nanosystems_ and you
didn't. Reasoning by analogy is no substitute for technical knowledge.
You might as well be writing crackpot revolutionary theories to some
weary physicist.
Likewise, appealing to a sense of future shock is no different than the
good old proofs that nobody would ever travel in space. Yes, Lyle,
everything can change in just one decade. It happens.
--
sentience@pobox.com Eliezer S. Yudkowsky
http://pobox.com/~sentience/AI_design.temp.html
http://pobox.com/~sentience/singul_arity.html
Disclaimer: Unless otherwise specified, I'm not telling you
everything I think I know.
This archive was generated by hypermail 2.1.5 : Fri Nov 01 2002 - 15:03:34 MST