"I AIN'T GOT NO SOUL!" (was: Re: Hyper-accelerating towards The Singularity)

Peter da Silva

unread,

Apr 23, 1998, 3:00:00 AM4/23/98

to

In article <m3k98g7...@windlord.Stanford.EDU>,
Russ Allbery <r...@stanford.edu> wrote:
>What does that question have to do with anything? There are a lot more
>things in this world besides number-crunching machines and mystery souls,
>Lee. For one thing, the brain is an analog device.

What's your basis for assuming that the analog brain couldn't theoretically
be simulated to any sufficiently detailed level on digital hardware?

(as a thought experiment, please, I realise that it's a hideously inefficient
way of creating an AI, though Greg Egan seems enamoured with it)

Observe followups.

--
This is The Reverend Peter da Silva's Boring Sig File - there are no references
to Wolves, Kibo, Discordianism, or The Church of the Subgenius in this document
| "Open": a warning label on a product or organization to let you know that |
| it is more proprietary and restrictive than its competitors. |

Peter da Silva

unread,

Apr 23, 1998, 3:00:00 AM4/23/98

to

Again, followups...

In article <353F876A...@neosoft.com>,
Herb Smith <hsm...@neosoft.com> wrote:
>I think you're not considering that there may be quantum level aspects
>to intelligence.

Of course there are. There are quantum level aspects to a light bulb. That
doesn't mean they can't be simulated.

I get the same squeenky feeling about people using "quantum level
aspects" to explain the inexplicable that Russ gets from people
confusing AI and science fiction.

>If in fact there are nonlocal aspects to sentience,
>it would provide an explanation for the various "wild talents" and
>so-called psi abilities, that are persistently reported, without
>delving into the "supernatural"

Miracles, angels, and honest politicians are also persistently reported.

In none of these cases is there enough evidence to indicate that there's
anything there that requires explanation.

Russ Allbery

unread,

Apr 23, 1998, 3:00:00 AM4/23/98

to

In net.general, Peter da Silva <pe...@taronga.com> writes:

> What's your basis for assuming that the analog brain couldn't
> theoretically be simulated to any sufficiently detailed level on digital
> hardware?

Quantum computer theory.

--
Russ Allbery (r...@stanford.edu) <URL:http://www.eyrie.org/~eagle/>

Mark Atwood

unread,

Apr 23, 1998, 3:00:00 AM4/23/98

to

pe...@taronga.com (Peter da Silva) writes:
>
> What's your basis for assuming that the analog brain couldn't theoretically
> be simulated to any sufficiently detailed level on digital hardware?
>

> (as a thought experiment, please, I realise that it's a hideously inefficient
> way of creating an AI, though Greg Egan seems enamoured with it)
>

I havnt read Digital City yet, but that wasn't the model in Disaspora.

The process of Yatima's birth was pretty detailed: the Citizens (and
the Gleisners as well) execute in a massivly multithreaded virtual
machine with a multidimentional address space. The address space is
initialized to zero, the mindseed (only a few hundred megabytes in
size) is copied into it, and execution begins. The opcode
specification was called "Shaper", and seemed mostly standard between
Poliss and even to the Gleisners' slower hardware. The opcodes
appeared to be 6 bits in size, and implicit operand, manipulating
values in the address "space" "around" them, and branching, and doing
"thread forks". It's kind of like N-dimentional "core war".

It was not the digital simulation of wetware electrochemistry, nor was
it DNA simulation.

The biggest bit of handwaving on his part, IMO, was how the Introdus
worked, especially since the nanotech was show to be very readily
manufactured, small, and able to be delivered without setup,
preparation, or aquiesence on the part of the subject.

I would bet that porting wetware to shaper would be an impossible
task, or at least much more difficult than synthysis of a "native"
software intellegence.

If hardware ever gets fast enough to do "digital simulation of wetware
electrochemistry", that may be the only direct upload path for us
fleshers.

I havnt decided what I believe in the "real world" re the argument
about "intellegence is only software" or "there is a
physical/quantum/magical process".

--
Mark Atwood | Thank you gentlemen, you are everything we have come to
m...@pobox.com | expect from years of government training. -- MIB Zed

Jeremy.

unread,

Apr 23, 1998, 3:00:00 AM4/23/98

to

In net.science.misc, Russ Allbery (r...@stanford.edu) said...

>In net.general, Peter da Silva <pe...@taronga.com> writes:
>
>> What's your basis for assuming that the analog brain couldn't
>> theoretically be simulated to any sufficiently detailed level on digital
>> hardware?
>

>Quantum computer theory.

But how do you rectify the results of QCT with the latest trends in
Probabilisitic Subatomic Computing? I must say I've never accepted QCT as
more than a fringe movement. Or course someone'll probably come along in
10 years and prove them both equivalent.

-J.
oh wait, I thought I was in net.bozo.science.misc. Sorry.

--
No lover's ever faithful / No contract truly signed...
Never take a stranger's advice / Never let a friend fool you twice...
Never stay a minute too long / Don't forget the best will go wrong.
--"Nobody's on Nobody's Side", _Chess_

Jim Kingdon

unread,

Apr 23, 1998, 3:00:00 AM4/23/98

to

> oh wait, I thought I was in net.bozo.science.misc. Sorry.

Oh, is _that_ where all the UFO posts which used to infest
sci.space.policy went?

It is almost as if sci.space.policy is calm again. It is back to the
overpopulation flamewar (which is a persistent pest, but not as
far-out and off-topic as UFOs).

Herb Smith

unread,

Apr 23, 1998, 3:00:00 AM4/23/98

to

Peter da Silva wrote:

>
> Herb Smith <hsm...@neosoft.com> wrote:
> >
> >If in fact there are nonlocal aspects to sentience,
> >it would provide an explanation for the various "wild talents" and
> >so-called psi abilities, that are persistently reported, without
> >delving into the "supernatural"
>
> Miracles, angels, and honest politicians are also persistently reported.
>
> In none of these cases is there enough evidence to indicate that there's anything there that requires explanation.

The "Miracle of the Sun" at Fatima, October 13, 1917, at the time
and spot announced by the children & witnessed by over 70,000 people
& reported in newspapers around the world?

SOMETHING happened there. I don't buy the Catholic explanation,
but I can't dismiss the evidence with a wave of my hand. At the
very least you have "mass hypnosis" capable of overcoming the
senses of confirmed atheists and debunkers there specifically to
scoff, and if you allow that theory, how may "historical" events
can also be the result of "mass hypnosis"?

Here you've either got children able to sense that SOMETHING
was going to happen months in the future, or a group of people,
fired by religious faith, able to produce either actual physical
events, or the perception of those events in distant observers. In
any of those events, I'd like an explanation for how it happened,
in a scientific framework.

--
Herb Smith - NeoSoft R&D NeoSoft, Inc.
hsm...@neosoft.com 713-968-5800
http://www.neosoft.com/~hsmith http://www.neosoft.com

Never trust a file that isn't ASCII.

Peter da Silva

unread,

Apr 23, 1998, 3:00:00 AM4/23/98

to

In article <m3u37k4...@windlord.Stanford.EDU>,

Russ Allbery <r...@stanford.edu> wrote:
>In net.general, Peter da Silva <pe...@taronga.com> writes:
>> What's your basis for assuming that the analog brain couldn't
>> theoretically be simulated to any sufficiently detailed level on digital
>> hardware?

>Quantum computer theory.

I don't believe in quantum computers. What it amounts to is using the
uncollapsed state of a system as a parallel processor that grows
exponentially in time. The resources required to keep the system coherent
through the process can also be expected to grow exponentially in time.

This is deep magic science fiction. Using it as an argument as to why
AIs, a much less controversial subject, are deep magic science fiction
is... well... pretty wild.

Peter da Silva

unread,

Apr 23, 1998, 3:00:00 AM4/23/98

to

In article <v6ogxsc...@colon.dev.ampersand.com>,
Mark Atwood <m...@pobox.com> wrote:

>pe...@taronga.com (Peter da Silva) writes:
>> What's your basis for assuming that the analog brain couldn't theoretically
>> be simulated to any sufficiently detailed level on digital hardware?

>> (as a thought experiment, please, I realise that it's a hideously inefficient

>> way of creating an AI, though Greg Egan seems enamoured with it)

>I havnt read Digital City yet, but that wasn't the model in Disaspora.

It was one of them. Yatima was not based on a human brain, but Orlando was.

The AIs in Permutation City are explicitly based on clinical simulations
of the human brain.

The "Jewels" in a couple of the stories in Axiomatic are a more heuristic
version of the same thing.

Peter da Silva

unread,

Apr 23, 1998, 3:00:00 AM4/23/98

to

In article <353FB9CA...@neosoft.com>,

Herb Smith <hsm...@neosoft.com> wrote:
>SOMETHING happened there. I don't buy the Catholic explanation,
>but I can't dismiss the evidence with a wave of my hand. At the
>very least you have "mass hypnosis" capable of overcoming the
>senses of confirmed atheists and debunkers there specifically to
>scoff, and if you allow that theory, how may "historical" events
>can also be the result of "mass hypnosis"?

Hypnosis itself is not an entirely well explained phenomenon, and
while it effects skeptics and believers alike its effects seem to
be very very dependent on what popular culture believes it should
be. At around the time you're talking about (I don't recall if it
was late 1800s or early 1900s when this effect showed up) hypnosis
itself was in a state of flux. While it never had this effect
before or after, everyone subjected to hypnosis would writhe in
the presence of magnets. Because they expected to, because everyone
knew that Mesmer's "animal magnetism" had an effect like that.

Or at least that's what the papers of the time reported.

So yes I DO accept the possibility that some kind of "mass hypnosis"
produced the reports. Or that people simply lied. It's interesting
that this sort of event has almost vanished from the world since
something resembling objective observers... the movie camera and its
descendents... has arrived on the scene.

r...@greenend.org.uk

unread,

Apr 24, 1998, 3:00:00 AM4/24/98

to

So I build a machine that can put individual molecules in the same
place they appear as in Peter's brain, and let it do its stuff I feed
the result the bloody supply etc. that it needs, wire it up to
something resembling eyes, ears, speech, et.c

Is the result an artificial intelligence? If not, what is it?

(Does it have a soul, if there are such things? If not, is that bad?)

Jim Kingdon

unread,

Apr 24, 1998, 3:00:00 AM4/24/98

to

> the Citizens (and the Gleisners as well) execute in a massivly
> multithreaded virtual machine with a multidimentional address
> space. The address space is initialized to zero, the mindseed (only a
> few hundred megabytes in size) is copied into it, and execution
> begins.

Hah! Emacs:

* Has a core a few hundred megabytes in size (:-))

* Has a variety of "citizens" (packages) executing within it

* Is a lot like N-dimensional "core war".

So let's just declare emacs an Artificial Intelligence and be done
with it. Or an Artificial Stupidity, I'm not picky.

Russ Allbery

unread,

Apr 24, 1998, 3:00:00 AM4/24/98

to

In net.science.misc, rjk <r...@greenend.org.uk> writes:

> So I build a machine that can put individual molecules in the same place
> they appear as in Peter's brain, and let it do its stuff I feed the
> result the bloody supply etc. that it needs, wire it up to something
> resembling eyes, ears, speech, et.c

> Is the result an artificial intelligence? If not, what is it?

A clone.

carpe carpus

unread,

Apr 24, 1998, 3:00:00 AM4/24/98

to

In article <wwvzphc...@sfere.greenend.org.uk>,

<r...@greenend.org.uk> wrote:
>So I build a machine that can put individual molecules in the same
>place they appear as in Peter's brain, and let it do its stuff I feed
>the result the bloody supply etc. that it needs, wire it up to
>something resembling eyes, ears, speech, et.c
>Is the result an artificial intelligence? If not, what is it?

Artifical, perhaps, only in the sense that you built the machine.

>(Does it have a soul, if there are such things? If not, is that bad?)

I don't know. Why don't you ask it if it has a soul, and see what it
says? It'll probably respond something tart in an Australian accent,
is my bet.

rone
--
I am ahead
I am advanced
I am the first mammal to wear pants

Russ Allbery

unread,

Apr 24, 1998, 3:00:00 AM4/24/98

to

Peter da Silva <pe...@taronga.com> writes:

> Russ Allbery <r...@stanford.edu> wrote:

>> Quantum computer theory.

> I don't believe in quantum computers. What it amounts to is using the
> uncollapsed state of a system as a parallel processor that grows
> exponentially in time. The resources required to keep the system
> coherent through the process can also be expected to grow exponentially
> in time.

Heh. Okay, I'll try to come up with something better than a throwaway
answer.

The answer is, quite simply, that we both have absolutely no idea how the
brain actually works and that we've been unable to build real computer
simulations of systems with only a minute fraction of the complexity.
We've been able to work out theoretically innumerable systems that we've
been totally unable to build, so there are even things we *understand*
(for some definition of understand) that we can't construct.

I find it quite plausible that we'll be able to eventually come close to
building a model of those parts of the brain that we can examine. At
which point we're going to run into several problems: It's very hard to
actually examine a functioning human brain (in the absence of tricorder
Star Trek technology), we don't really know what happens when it dies and
how that might change things, and we're dealing with analog technology
that is probably quite vulnerable to the Butterfly Effect. In other
words, getting the fortieth decimal place wrong in a simulation could
destroy the entire thing.

Even if we manage to build an exact simulation of the human brain, that
doesn't mean we necessarily *understand how it works*. People build
replicas of things they don't understand all the time. And, finally, none
of this has anything at all to do with AIs, since that's a discussion of
how to duplicate people. Unless you consider cloning humans to be
creation of artificial intelligence.

> This is deep magic science fiction.

QC is not deep magic science fiction. QC is real life graduate computer
science research.

> Using it as an argument as to why AIs, a much less controversial
> subject,

Oh, c'mon, Peter. SF-level AIs most certainly are as controversial as
QC. They're as controversial as warp drives. They're just one of the
controversial science subjects where science fiction as a field has
basically decided they don't care what real scientists have to say on the
subject except insofar as it generates another wildly implausible
handwaving explanation for tech they want to use.

Tony Finch

unread,

Apr 24, 1998, 3:00:00 AM4/24/98

to

Jim Kingdon <kin...@panix3.panix.com> wrote:
>
>Hah! Emacs:
>
>* Has a core a few hundred megabytes in size (:-))
>
>* Has a variety of "citizens" (packages) executing within it
>
>* Is a lot like N-dimensional "core war".

Now you've got me thinking of a way to get my nascent rcp package to
have a fight with ange-ftp and boot it out of the system....

Tony.
--
"Gently break the heart of a young lettuce."

Peter da Silva

unread,

Apr 24, 1998, 3:00:00 AM4/24/98

to

In article <rone.FriApr24095...@ennui.org>,

carpe carpus <rone+...@ennui.org> wrote:
>I don't know. Why don't you ask it if it has a soul, and see what it
>says? It'll probably respond something tart in an Australian accent,
>is my bet.

It says "I don't even have any bloody feet, you idiot".

Peter da Silva

unread,

Apr 24, 1998, 3:00:00 AM4/24/98

to

In article <m390ov3...@windlord.Stanford.EDU>,

Russ Allbery <r...@stanford.edu> wrote:
>The answer is, quite simply, that we both have absolutely no idea how the
>brain actually works and that we've been unable to build real computer
>simulations of systems with only a minute fraction of the complexity.

I don't agree that we have no idea how the brain works. We're not at the
Newtonian Mechanics level yet, let alone Einstein's or Heisenberg's models
or the GUT level that will let us build our own brainlets, but we do have
quite a lot of ideas about how parts of the brain work, and some of them
even stand up reasonably well to experimentation.

I don't expect AIs in my lifetime or my children's lifetime, but I find the
idea that there's some fundamental impossibility in building one quite
bizzarre.

>We've been able to work out theoretically innumerable systems that we've
>been totally unable to build, so there are even things we *understand*
>(for some definition of understand) that we can't construct.

We can't construct lots of things we have quite a good understanding of.

That doesn't mean we have no idea how they work.

>I find it quite plausible that we'll be able to eventually come close to
>building a model of those parts of the brain that we can examine. At
>which point we're going to run into several problems: It's very hard to
>actually examine a functioning human brain (in the absence of tricorder
>Star Trek technology), we don't really know what happens when it dies and
>how that might change things, and we're dealing with analog technology
>that is probably quite vulnerable to the Butterfly Effect. In other
>words, getting the fortieth decimal place wrong in a simulation could
>destroy the entire thing.

It could. However the human brain has proven itself quite robust in the
face of severe damage.

>Even if we manage to build an exact simulation of the human brain, that
>doesn't mean we necessarily *understand how it works*.

No, but it does mean that *how it works* doesn't involve magic.

>People build
>replicas of things they don't understand all the time. And, finally, none
>of this has anything at all to do with AIs, since that's a discussion of
>how to duplicate people. Unless you consider cloning humans to be
>creation of artificial intelligence.

If you could do it it would stand as a proof that human level intelligence
doesn't require any facilities that a digital computer couldn't provide.

It would prove that AI was possible. It wouldn't prove that it was easy nor
that it could be accomplished within the lifetime of our civilization, but
it would get rid of the deep magic.

>> This is deep magic science fiction.

>QC is not deep magic science fiction. QC is real life graduate computer
>science research.

I don't see a contradiction between these two statements. Physicists do work
all the time on things they have no expectation of ever being able to build.
Wheeler's about as serious a physicist as you can find yet he put his name
on a paper describing a multiple worlds model that he's STATED he considers
fantasy.

It satisfies the mathematics. That doesn't mean it's real. And QC requires
something damn close to it to BE real to work.

>> Using it as an argument as to why AIs, a much less controversial
>> subject,

>Oh, c'mon, Peter. SF-level AIs most certainly are as controversial as
>QC.

QC requires new physics. SF-level AIs are simply very very difficult.

>They're as controversial as warp drives.

Nonsense. We don't even have a physical theory to suggest that warp drives
might be possible. We have never observed an object travelling faster than
light, on any scale. The closest we've come is the Aspect experiment and that
effect can't be used to transmit information let alone move starships.

We have working examples of intelligences.

Russ Allbery

unread,

Apr 24, 1998, 3:00:00 AM4/24/98

to

Peter da Silva <pe...@taronga.com> writes:

> So what is "artificial intelligence" to you if a human brain emulated on
> a digital computer doesn't qualify?

The original post said a machine that "can put individual molecules in the
same places that they are in Peter's brain." To me, that isn't am
emulation, but rather is a full-out clone.

Shelton Garner

unread,

Apr 24, 1998, 3:00:00 AM4/24/98

to

On 24 Apr 1998 09:57:43 -0700, Russ Allbery <r...@stanford.edu> wrote:

>In net.science.misc, rjk <r...@greenend.org.uk> writes:
>

>> So I build a machine that can put individual molecules in the same place

>> they appear as in Peter's brain, and let it do its stuff I feed the
>> result the bloody supply etc. that it needs, wire it up to something
>> resembling eyes, ears, speech, et.c
>
>> Is the result an artificial intelligence? If not, what is it?
>

>A clone.

Hmmm, what if we were able to create "enhanced biologically manmade
people" (a la "Blade Runner")...let's call them "cyborgs" for want of
better word...I'd call them AIs, esp if they had a few computer chips
in their heads.

lee

L. Shelton Bumgarner -- Keeper of the Great Renaming FAQ
Nattering Nabob of Narcissism * http://www.nottowayez.net/~leebum/
ICQ#: 9393354

Russ Allbery

unread,

Apr 25, 1998, 3:00:00 AM4/25/98

to

Peter da Silva <pe...@taronga.com> writes:

> Russ Allbery <r...@stanford.edu> wrote:

>> The original post said a machine that "can put individual molecules in
>> the same places that they are in Peter's brain." To me, that isn't am

>> emulation, but rather is a full-out clone.

> In the context if the two messages previous to this one it's emulations
> of the molecules.

That would be artificial human intelligence, then.

Russ Allbery

unread,

Apr 25, 1998, 3:00:00 AM4/25/98

to

Peter da Silva <pe...@taronga.com> writes:

> I don't expect AIs in my lifetime or my children's lifetime, but I find
> the idea that there's some fundamental impossibility in building one
> quite bizzarre.

I don't know that that's really what I'm arguing. I personally tend to
consider "apparently impossible" and "requiring such advanced tech as to
be impossible to predict or anticipate within any sort of reasonable time
frame" to be roughly equivalent, since I really don't consider much at all
to be absolutely *impossible.* Attempting to prove a negative and all
that.

I consider construction of an artificial intelligence to be in the same
category as warp travel, time travel, moving Venus, or building a Dyson
sphere. All of those have some sort of scientific basis in one theory or
another, all of those have been credibly argued to be effectively
impossible, and all of those people have similarly argued *are* possible.
All of those are sufficiently far beyond the grasp of mankind currently or
in the forseeable future that by the time a society is capable of actually
making plans to do any of them, I find it unlikely that the society will
be recognizeably human to us.

(Similarly, I think a lot of great science fiction has been written about
all of those topics, and certainly don't want people to stop.)

> I wrote:
>> It's very hard to actually examine a functioning human brain (in the
>> absence of tricorder Star Trek technology), we don't really know what
>> happens when it dies and how that might change things, and we're
>> dealing with analog technology that is probably quite vulnerable to the
>> Butterfly Effect. In other words, getting the fortieth decimal place
>> wrong in a simulation could destroy the entire thing.

> It could. However the human brain has proven itself quite robust in the
> face of severe damage.

True. Although in most cases this has been based on a model that was
*already* working, plus it's somewhat questionable what precisely one
means by "working." (This is one of the problems with the whole field of
artificial intelligence, when it strays outside of the fairly limited
application in applied CS. What *is* intelligence? The field can't even
really define its own goal.) I have quite a bit of confidence that
eventually man will be able to build a brain capable of performing the
rough equivalents of the day-to-day processing that the average extremely
simple animal does. I think there's a quantum leap inbetween that and the
human brain.

>> Even if we manage to build an exact simulation of the human brain, that
>> doesn't mean we necessarily *understand how it works*.

> No, but it does mean that *how it works* doesn't involve magic.

Actually, it doesn't even clearly imply that, depending on what you mean
by magic. See Orson Scott Card's thoughts on the subject in _Xenocide_,
which basically amount to a slightly saner thetan theory. (The executive
summary is that a theory involving energy beings from another plane of
existence that are attracted to certain constructed structures in this
universe is just as plausible.) I should point out that I personally have
very little respect for the handwaving sorts of explanations in _Xenocide_
and thought that was one of the weaker parts of the book, but it isn't
clearly ruled out by the construction of a simulation of a human brain.

FWIW, I honestly have no idea if there's any "magic" involved in the
working of the brain or not; it doesn't much matter to me either way.

>> People build replicas of things they don't understand all the time.
>> And, finally, none of this has anything at all to do with AIs, since
>> that's a discussion of how to duplicate people. Unless you consider
>> cloning humans to be creation of artificial intelligence.

> If you could do it it would stand as a proof that human level
> intelligence doesn't require any facilities that a digital computer
> couldn't provide.

That it would.

> It would prove that AI was possible.

Actually, no, it wouldn't (if by AI you mean created intelligence distinct
from simulated human intelligence), since that postulates the unproven
theory that there exist, at least as implementable possibilities, forms of
intelligence other than human intelligence. :) We don't have enough
understanding of the subject to even know if there's anything "special"
about the way in which humans in particular think that in the end will
imply that only very similar thinking patterns are possible in intelligent
beings.

>> QC is not deep magic science fiction. QC is real life graduate
>> computer science research.

> I don't see a contradiction between these two statements.

Mm. Point. Fair enough.

>> They're as controversial as warp drives.

> Nonsense. We don't even have a physical theory to suggest that warp
> drives might be possible.

Actually, we have a few that wander around. Although you may have a more
stringent requirement for calling something a "physical theory" than I do.

> We have never observed an object travelling faster than light, on any
> scale. The closest we've come is the Aspect experiment and that effect
> can't be used to transmit information let alone move starships.

> We have working examples of intelligences.

This is true.

(The other side of this conversation that I think people miss is that it
isn't at all clear to me why we would *want* to build simulations of human
intelligence even if we could, except to prove a few points. There are
considerably simpler methods of manufacturing human beings.)

Ryan Tucker

unread,

Apr 26, 1998, 3:00:00 AM4/26/98

to

On 26 Apr 1998 07:11:42 -0500, Peter da Silva <pe...@taronga.com> spewed:
> 2. Humans don't last very long.
>
> With careful backups and maintainance, there's no reason
> a non-organic human couldn't last indefinitely.

Brings new meaning to "You DO have a backup... right?" -rt

--
Ryan Tucker <rtucker...@ttgcitn.com>
http://www.ttgcitn.com/~rtucker/

Kai Henningsen

unread,

Apr 26, 1998, 3:00:00 AM4/26/98

to

r...@stanford.edu (Russ Allbery) wrote on 25.04.98 in <m3yawtr...@windlord.Stanford.EDU>:

> I don't know that that's really what I'm arguing. I personally tend to
> consider "apparently impossible" and "requiring such advanced tech as to
> be impossible to predict or anticipate within any sort of reasonable time
> frame" to be roughly equivalent, since I really don't consider much at all
> to be absolutely *impossible.* Attempting to prove a negative and all
> that.

And I personally consider those about as far apart as can be. Let alone
there's a serious difference between believing something and proving it,
and further that anyone involved with science should know that in certain
situations, you _can_ prove a negative (as long as you can take some basic
assumptions as given, or while in math).

> I consider construction of an artificial intelligence to be in the same
> category as warp travel, time travel, moving Venus, or building a Dyson
> sphere.

Dyson sphere, yes. The others, no. (And Venus is no mainly because a
planet seems too fragile. I could be wrong on that one. Though anyway I'd
prefer a plan that makes it hospitable in place, which is definitely not
impossible.)

We have _no_ basis for believing that warp or time travel is possible. We
do have such a basis for AIs and Dyson spheres.

(We do have some theories that point to time travel in certain unlikely
circumstances. However, we have no idea at all if those theories describe
reality.)

> All of those have some sort of scientific basis in one theory or
> another, all of those have been credibly argued to be effectively
> impossible, and all of those people have similarly argued *are* possible.

I'll state flat out that no argument about AI being impossible has ever
passed my credibility filter. Ultimately, *all* of them appeal to religion
or fantasy science. I'll believe them when I believe an existance proof
for God, or (as an easier condition) when hell freezes over.

I've not even seen someone try to prove Dyson spheres are impossible.

> All of those are sufficiently far beyond the grasp of mankind currently or
> in the forseeable future that by the time a society is capable of actually
> making plans to do any of them, I find it unlikely that the society will
> be recognizeably human to us.

Probably true for all *except* AI.

> What *is* intelligence? The field can't even
> really define its own goal.)

"I know it when I see it" :-)

> eventually man will be able to build a brain capable of performing the
> rough equivalents of the day-to-day processing that the average extremely
> simple animal does.

The computer you wrote than on is probably far more advanced than some
animal brains out there. Face it; simple animals just don't have that much
brain power. It's not hard to do, say, the equivalent of a fly.

> I think there's a quantum leap inbetween that and the
> human brain.

Sure. That's what we're working on right now.

I think Hogan (in the book I mentioned in another article) was exactly
right in his assumption where "real" AI will come from: from a system
designed to solve our problems for us. A system that you can tell "I need
X", and it finds some way to do X, by putting together a lot of facts that
we fed it.

And the main reason why we're so far from "real" AI today is the gigantic
amount of facts we take for granted in human brains. All of these need to
be predigested and fed into computers. And then computers need to be fast
enough to actually do something useful with those facts in reasonable
time.

There's absolutely nothing impossible about any of this. It's just hard.
And it's useful enough to practically guarantee that it gets built, some
day.

Incidentally, five years ago, I'd not have expected computers to be as
good at this as they are today. It's going far faster than I expected.
Even something as basically dumb as Altavista is something I didn't expect
to see for at least a decade longer.

(Is it the old cathedral-vs.-bazaar thing again?)

> > It would prove that AI was possible.
>
> Actually, no, it wouldn't (if by AI you mean created intelligence distinct
> from simulated human intelligence), since that postulates the unproven
> theory that there exist, at least as implementable possibilities, forms of
> intelligence other than human intelligence. :)

That one has existance proofs in my book. I happen to believe, from the
evidence available, that certain sea-inhabiting mammals are quite similar
in intelligence magnitude to us. (Also extremely different in specifics,
so much so that they essentially qualify for SF aliens, with all the
communication problems that entails.)

> (The other side of this conversation that I think people miss is that it
> isn't at all clear to me why we would *want* to build simulations of human
> intelligence even if we could, except to prove a few points. There are
> considerably simpler methods of manufacturing human beings.)

Well, should we ever find a way to simulate specific individuals, I could
see a point :-)

Otherwise, I'd be much more interested in _different_ intelligences - and
those are probably far simpler to come by anyway.

Kai
--
http://www.westfalen.de/private/khms/
"... by God I *KNOW* what this network is for, and you can't have it."
- Russ Allbery (r...@stanford.edu)

Russ Allbery

unread,

Apr 26, 1998, 3:00:00 AM4/26/98

to

Peter da Silva <pe...@taronga.com> writes:

> Russ Allbery <r...@stanford.edu> wrote:

>> I don't know that that's really what I'm arguing.

> It sure seems like it.

That would seem to indicate that we're having trouble communicating with
each other, then, wouldn't it?

[...]

> There's lots of different kinds of impossibles.

> There's "We know basically how to do this, but we don't have the resources".
> There's "We know this is possible, but we don't know how to do it".
> And there's "We don't even know if this universe's physics allows for it".

Okay, fair enough. You're right, I'm being sloppy. I largely agree with
you and would put AIs in category two, for a sufficiently broad definition
of AI. If you're speaking specifically about simulating intelligence in a
digital computer, I think that it's unclear if that belongs in category
two or category three at this point; it's entirely possible that
intelligence requires effects not reproducible in a digital device. Our
one class of examples of intelligence are all analog.

> And, frankly, your quantum computers fit in here.

The point I was trying to make in bringing up quantum computers is the
point above about whether we can simulate intelligence in a digital
computer. We know that there are *other* sorts of computing at least
theoretically possible which can do things that digital computers cannot,
and we don't know enough about how the brain works to know whether or not
it involves any of them.

>> (This is one of the problems with the whole field of artificial
>> intelligence, when it strays outside of the fairly limited application
>> in applied CS. What *is* intelligence? The field can't even really
>> define its own goal.)

> Only if you think it needs to have a single unique goal. I'm sure glad
> physics doesn't, or else we might be running around still trying to
> satisfy the goals extablished by the Alchemists.

Physics can at least provide you with a definition of the terms that make
up its name. So can applied AI; theoretical AI has more problems there.

>> I think there's a quantum leap inbetween that and the human brain.

> I don't. I don't see that there's any evidence to imply that the
> difference between human level intelligence and animal intelligence is
> anything but a difference in quantity.

I think this is the major open question in the field. Until someone
actually answers it conclusively one way or the other, we'll have to agree
to disagree on this one.

> We don't really do anything differently from them, we just do more of
> it.

That's a rephrasing of the same open question. It isn't clear to me that
this is true.

>>> No, but it does mean that *how it works* doesn't involve magic.

>> Actually, it doesn't even clearly imply that, depending on what you
>> mean by magic. See Orson Scott Card's thoughts on the subject in
>> _Xenocide_, which basically amount to a slightly saner thetan theory.
>> (The executive summary is that a theory involving energy beings from
>> another plane of existence that are attracted to certain constructed
>> structures in this universe is just as plausible.)

> You're *really* reaching. Card is a nut even by SF writer's standards.

You're making flat statements about what something proves. You should
know better than that. :)

>> FWIW, I honestly have no idea if there's any "magic" involved in the
>> working of the brain or not; it doesn't much matter to me either way.

> You seem to be working pretty hard to find one I can't disprove.

I like discussing the subject.

> That's the sort of argument I'd expect from a creationist.

That's nice. And you're a Nazi. Can we stop insulting each other now and
talk about ideas? Creationism is pseudo-scientific bunk.

> And if it's true then that means that the AI's we build will be very
> much like us, which is a very interesting result in and of itself. It
> would certainly make Star Trek more beleivable.

The political and social structure in Star Trek is a hell of a lot more
unbelievable than the AIs. :)

>> Actually, we have a few that wander around. Although you may have a
>> more stringent requirement for calling something a "physical theory"
>> than I do.

> I suspect so, given the mormon-filtered dianetics you were postulating
> up there.

Oh, good grief. That wasn't a physical theory. That was a piece of
science fiction that I said was handwaved and badly written. There are a
variety of theories in physics that are at about the same level as QC is
that allow for the theoretical possibility of various sorts of warp
drives. These are the sorts of theories I was referring to earlier when I
said that science fiction writers paid attention to them to get more
handwaving sorts of justifications. Krazinov Tubes, for example. Stuff
like that where the math works out or might work out, but that we have no
idea whether or not it works.

I'm guessing that your inclusion of the word "physical" in "physical
theory" was designed to rule those out, but I don't know the precise
definitions that you're using.

>> (The other side of this conversation that I think people miss is that
>> it isn't at all clear to me why we would *want* to build simulations of
>> human intelligence even if we could, except to prove a few points.
>> There are considerably simpler methods of manufacturing human beings.)

> Lots of reasons:

> 1. Human beings are pretty fragile. We reuire basically a
> whole planet's worth of life support over the long term.

> 2. Humans don't last very long.

These two assume digital simulations, not biological simulations or
constructed humans.

> 3. If we understand what's necessary to be a human, then we can
> build structures that perform that into other organic life
> forms.

It's not clear to me why we would want to do this, given the expense
involved. It's also not clear to me that biology is sufficiently
adaptable as to allow "fast evolution" of that sort.

> 4. If you can build a simulation of a brain, you can build a
> simulation of part of a brain.

> Traumatic brain damage could be repaired. Certain kinds of
> dysfunctional behaviour could be worked around.

This one I wholeheartedly agree with, and furthermore expect to see *far*
sooner than actual AI. This doesn't require AI; this just requires enough
understanding of the general function to build a prosthetic. We could
build workable prosthetic limbs long before we really understood how
biological limbs worked.

> I suspect that not all of these are equally probable, and things like #3
> are pretty far-fetched, but surely at least some of these are worthwhile
> if they could come out of the research.

#4 is the most immediately useful and also the most likely. #3 I'd rather
not even speculate on, given that I have very, very little background in
biology and not much ability to speak knowledgably about it beyond being a
little skeptical. #1 and #2 strike me as hard, possibly Very Hard,
problems.

Peter da Silva

unread,

Apr 26, 1998, 3:00:00 AM4/26/98

to

In article <m3k98cr...@windlord.Stanford.EDU>,

Russ Allbery <r...@stanford.edu> wrote:
>Okay, fair enough. You're right, I'm being sloppy. I largely agree with
>you and would put AIs in category two, for a sufficiently broad definition
>of AI. If you're speaking specifically about simulating intelligence in a
>digital computer, I think that it's unclear if that belongs in category
>two or category three at this point; it's entirely possible that
>intelligence requires effects not reproducible in a digital device. Our
>one class of examples of intelligence are all analog.

It's possible, yes, and it's possible that at some point our space probes are
going to crash on Brin's crystal spheres. There's even evidence that some such
structure exists: the Great Silence. I consider these two possibilities to be
within an order of magnitude of the same probability.

>> And, frankly, your quantum computers fit in here.

>The point I was trying to make in bringing up quantum computers is the
>point above about whether we can simulate intelligence in a digital
>computer. We know that there are *other* sorts of computing at least
>theoretically possible which can do things that digital computers cannot,

QC can't do anything digital computers can't. They can just do it faster.
Now it's possible, I guess, that the human brain involves structures that
collapse the state vector under controlled conditions to execute NP-hard
computations in finite time. But if that's the case then QC must actually
be relatively easy to make, and what we call "conventional" computers in
50 years will all include quantum components...

If you want a science fiction reference, toss in Greg Egan's "Quarantine" at
this point. I didn't find it particularly convincing.

>and we don't know enough about how the brain works to know whether or not
>it involves any of them.

What we do know about the brain doesn't provide any evidence that it does.

>>> I think there's a quantum leap inbetween that and the human brain.

>> I don't. I don't see that there's any evidence to imply that the
>> difference between human level intelligence and animal intelligence is
>> anything but a difference in quantity.

>I think this is the major open question in the field.

It's an open question because there's a lot of emotional baggage tied up
in there being some qualitative difference between humans and animals and
there's no conclusive evidence, yet, that there isn't.

There's even less evidence that there is.

>You're making flat statements about what something proves. You should
>know better than that. :)

What level of proof are you demanding?

Russ Allbery

unread,

Apr 26, 1998, 3:00:00 AM4/26/98

to

Peter da Silva <pe...@taronga.com> writes:

> QC can't do anything digital computers can't. They can just do it
> faster.

Heh. Amounts to the same thing oftentimes in computing. :) Quadratic
solutions vs. exponential solutions can very easily make the difference
between "answer in a day or two" and "answer sometime after the the sun
has burned out."

> What we do know about the brain doesn't provide any evidence that it
> does.

This seems to be a common theme throughout the rest of your message, so
rather than responding to it each time it comes up, I'll address it
generally. I am, saying more probably about my personality than about my
understanding of scientific theories, generally a cautious pragmatist when
it comes to things like this. I certainly am *not* a futurist. It's my
inclination to believe most problems to be hard until I see evidence to
the contrary, not the other way around.

I also have a healthy dose of exposure to current research in AI, which is
rather dramatically far away from creating intelligent computers. It is,
however, producing a lot of nice results in terms of leveraging off some
of our understanding of how humans think to get new tools for solving
problems.

> It's an open question because there's a lot of emotional baggage tied up
> in there being some qualitative difference between humans and animals
> and there's no conclusive evidence, yet, that there isn't.

> There's even less evidence that there is.

Um, actually, given that there's no conclusive evidence either way, it
would be an open question regardless. That's sort of the definition of an
open question. :)

Yes, a lot of the reason why it's such a *hyped* open question is because
of the emotional baggage. That's the same reason why people like Lee keep
hyping the prospect of AIs, and it's very similar to the rampant idiocy
out there concerning cloning. (I really have to wonder what precisely
people will think will happen if we clone a human being, apart from there
being one more human being.) I admit to being rather annoyed at the hype
on both sides.

> I'm sorry, but you were bringing up pseudo-scientific bunk to support
> your argument. That was silly, irrelevant, and insulting.

Oh please.

Peter, you said that construction of an AI would show that there's no
magic involved in intelligence. I'm sorry, but depending on what you mean
by "magic," it would do nothing of the sort. That particular question is
something that you've already made up your mind about, as have the people
on the other side, and the existence or lack thereof of AIs are not going
to change anyone's mind. I presented an example of why that is.

It had nothing to do with supporting my argument. It was an aside. Cope.
Or don't cope, I really don't care. But I really can do without the
gratuitous insults, thank you.

> I'm talking about the fact that virtually all aliens look and act human.
> If human intelligence was the only possible kind, that would make sense.

Ah, yeah. Good point.

>> 1. Human beings are pretty fragile. We reuire basically a
>> whole planet's worth of life support over the long term.
>> 2. Humans don't last very long.

> Well, they imply simulations where the "software" is independant of the

> "hardware", whether the "hardware" is a quantum computer, a digital
> computer, or an analog neural network.

(#1 mostly implies changing the life support structure to something that
functions on energy rather than on the various biological products of it.
You and I are mostly talking about #2.)

With the additional component that the software is copyable. There's a
lot more than simple AI involved there; ability to store and restore
global state is something that no intelligence we have any experience with
is capable of. It's certainly appealing; it's also in the realm of "we
have no idea if this is even possible" since we have no working examples.
People trying to attain this are leveraging off existing characteristics
of the media in which they're attempting to construct simulations.

#1 seems addressable in a lot of ways, both through creating life support
structures that require different sorts of things than what humans need
and possibly through changing what humans need. #2 is the tricky one,
both scientifically and ethically. (Most of the systems under which one
can create backups imply the ability to have multiple instances of the
same intelligence structure in operation simultaneously, which starts
making people extremely nervous.)

> That's sort of implied to me by the word "simulation". At least can you
> see why that sort of simulation would be attractive?

Actually, I can easily see a simulation in which the state of the
simulation is not extractable and restoreable.

And sure. You're right, I should amend my previous statement to wondering
why we'd care to make biological simulations or replicas of humans. If we
can change the basic structure, that may be useful in a lot of ways.

>>> 3. If we understand what's necessary to be a human, then we can
>>> build structures that perform that into other organic life
>>> forms.

>> It's not clear to me why we would want to do this, given the expense
>> involved.

> It wouldn't be clear to Thomas Jefferson why the United States would
> want multiple fleets of airplanes carrying hundreds of people at
> hundreds of miles an hour, given that the expense involved in building
> one would be beyond the reach of the world's resources in 1770.

> Arguing that some operation would be too expensive, in the future, isn't
> very convincing.

I think a lot of people miss how inherently cheap human beings are to
manufacture and maintain, provided that you have a reasonable amount of
time to wait. It's likely to be more cost-effective for most applications
to just make more humans. It's not an absolute statement about expense;
it's a relative one. As construction of AIs gets cheaper, so will
maintenance of human life. (And creation of human life is damn near
free.)

A lot of that depends on the advantages that such "animal intelligences"
(bad name, since we're animals too, but it gets the idea across) would
have over humans. And whether the additions would be inherited, in which
case a lot of the resource argument is moot since all you need is a
breeding pair.

Plus, there isn't just resource expense; there's also political and social
expense. Which means that it's likely to happen if the theory gets worked
out anyway, as soon as the resource expense gets within the reach of a
rich eccentric, but the upheaveal resulting is likely to be nasty.

>> It's also not clear to me that biology is sufficiently adaptable as to
>> allow "fast evolution" of that sort.

> It's not "fast evolution". We already have "fast evolution", in the
> form of artificial selection. It's just more involved prosthetics.

It's not a prosthesis. A prosthesis by definition replaces a capability
that would naturally be there but is missing. Adding or increasing
intelligence is adding something that isn't already there, and the parts
that *are* already there have to be able to use the new part. With a
prosthesis, you know that they can since they would normally have the
thing you're replacing.

And by "fast evolution," I mean doing something to increase intelligence
over the short term (one or two generations) rather than the long term
involved in any sort of evolutionary approach, presumably including what
you mean by artificial selection.

>> This one I wholeheartedly agree with, and furthermore expect to see
>> *far* sooner than actual AI. This doesn't require AI; this just
>> requires enough understanding of the general function to build a
>> prosthetic. We could build workable prosthetic limbs long before we
>> really understood how biological limbs worked.

> Depending on the level of "understanding" you're talking about. Building
> workable prosthetic limbs is one way you get part of the information
> needed to understand how limbs work. Similarly, neural prosthetics are
> on the road to AI whether or not that's the short-term goal.

Agreed.

Peter da Silva

unread,

Apr 27, 1998, 3:00:00 AM4/27/98

to

In article <m3yawrr...@windlord.Stanford.EDU>,

Russ Allbery <r...@stanford.edu> wrote:
>This seems to be a common theme throughout the rest of your message, so
>rather than responding to it each time it comes up, I'll address it
>generally. I am, saying more probably about my personality than about my
>understanding of scientific theories, generally a cautious pragmatist when
>it comes to things like this. I certainly am *not* a futurist. It's my
>inclination to believe most problems to be hard until I see evidence to
>the contrary, not the other way around.

I didn't say it wasn't a hard problem. I just don't think the fact that it's
a hard problem means that it's not going to be cracked.

>I also have a healthy dose of exposure to current research in AI, which is
>rather dramatically far away from creating intelligent computers. It is,
>however, producing a lot of nice results in terms of leveraging off some
>of our understanding of how humans think to get new tools for solving
>problems.

Yeh, you're at the Leonardo da Vinci point.

>> It's an open question because there's a lot of emotional baggage tied up
>> in there being some qualitative difference between humans and animals
>> and there's no conclusive evidence, yet, that there isn't.

>> There's even less evidence that there is.

>Um, actually, given that there's no conclusive evidence either way, it
>would be an open question regardless. That's sort of the definition of an
>open question. :)

There are some open questions that no sane scientist is willing to bother
betting on (the Face on Mars is a Sign of an Ancient Civilization, before
the latest photos), and there are some open questions they go "ah, well,
that's an open question". Because there's all this non-technical baggage
that gets them hassled if they tell you what they really think unless they're
hard-nosed pragmatists like Skinner.

Remember, there's never really conclusive evidence of anything. There are
just greater and lesser probabilities. The possibility that humans have any
qualitative differences in brain structure from every other animal on this
planet, well, I would have given the Face on Mars better odds.

>Yes, a lot of the reason why it's such a *hyped* open question is because
>of the emotional baggage. That's the same reason why people like Lee keep
>hyping the prospect of AIs, and it's very similar to the rampant idiocy
>out there concerning cloning.

I have no idea what Lee's saying about AIs and I could care less. I've had
him killfiled for months. The only person I'm discussing this with is you.

>> I'm sorry, but you were bringing up pseudo-scientific bunk to support
>> your argument. That was silly, irrelevant, and insulting.

>Peter, you said that construction of an AI would show that there's no

>magic involved in intelligence. I'm sorry, but depending on what you mean
>by "magic," it would do nothing of the sort.

"Magic" means "requires new physics to work".

>It had nothing to do with supporting my argument. It was an aside. Cope.
>Or don't cope, I really don't care. But I really can do without the
>gratuitous insults, thank you.

You really have no idea what you said, then, that pissed me off enough to
start going on about creationists? Forget it then.

>>> 1. Human beings are pretty fragile. We reuire basically a
>>> whole planet's worth of life support over the long term.
>>> 2. Humans don't last very long.

>(#1 mostly implies changing the life support structure to something that

>functions on energy rather than on the various biological products of it.

More or less, yes.

>You and I are mostly talking about #2.)

>With the additional component that the software is copyable. There's a
>lot more than simple AI involved there; ability to store and restore
>global state is something that no intelligence we have any experience with
>is capable of.

We've never had an intelligence that we can build a separate pathway for
saving and restoring that state into either.

Yes, it's certainly possible that there's some "magic" involved (see above)
that prevents it, but (a) it's unlikely, and (b) it's irrelevant: you asked
me "why do people see AIs as attractive" not "what will we be able to do with
AIs". I'm answering that question.

>#1 seems addressable in a lot of ways, both through creating life support
>structures that require different sorts of things than what humans need
>and possibly through changing what humans need. #2 is the tricky one,
>both scientifically and ethically. (Most of the systems under which one
>can create backups imply the ability to have multiple instances of the
>same intelligence structure in operation simultaneously, which starts
>making people extremely nervous.)

Technology is always making people nervous, and sometimes they have a good
reason to be nervous, but that's a side issue. Again, I'm talking about "why
do people see AIs as attractive".

>And sure. You're right, I should amend my previous statement to wondering
>why we'd care to make biological simulations or replicas of humans.

Where'd that come from, though?

>> Arguing that some operation would be too expensive, in the future, isn't
>> very convincing.

>I think a lot of people miss how inherently cheap human beings are to
>manufacture and maintain, provided that you have a reasonable amount of
>time to wait. It's likely to be more cost-effective for most applications
>to just make more humans. It's not an absolute statement about expense;
>it's a relative one. As construction of AIs gets cheaper, so will
>maintenance of human life. (And creation of human life is damn near
>free.)

This is a side issue, but I'll toss in a paragraph on it:

I'm talking about a civilization that's as much richer than we are than we
are richer than the one Benjamin Franklin lived in. People do things for
fun today that would have involved a major mobilization of the colonies in
1770. Lewis and Clark was enough of a Big Deal that many people simply didn't
believe they did it. Lewis committed suicide over that. My grandmother took
a trip to Antarctica for a vacation. My grandchildren will, assuming we don't
have a catastrophic breakdown, visit the moon with about as much fanfare.

>A lot of that depends on the advantages that such "animal intelligences"
>(bad name, since we're animals too, but it gets the idea across) would
>have over humans.

That's one of the things we'd want to find out, yes?

Anyway you *can* see why people would consider them desirable, even if you
don't see them as being likely. Which was the point.

>Plus, there isn't just resource expense; there's also political and social
>expense. Which means that it's likely to happen if the theory gets worked
>out anyway, as soon as the resource expense gets within the reach of a
>rich eccentric, but the upheaveal resulting is likely to be nasty.

Absolutely. The science fiction story I've been trying to write depends on it.

[skip the fast evolution/prosthetics argument, it's over words]

Russ Allbery

unread,

Apr 28, 1998, 3:00:00 AM4/28/98

to

Peter da Silva <pe...@taronga.com> writes:

> Remember, there's never really conclusive evidence of anything. There
> are just greater and lesser probabilities. The possibility that humans
> have any qualitative differences in brain structure from every other
> animal on this planet, well, I would have given the Face on Mars better
> odds.

Well, there's conclusive proof in the form of construction. I know that
Usenet is possible because we've constructed Usenet. But yes, I
understand what you mean.

> You really have no idea what you said, then, that pissed me off enough
> to start going on about creationists? Forget it then.

Nope. No idea.

I do know what pissed me off enough to respond the way that I did. Every
time I get started on AIs (yeah, this isn't a new topic with me), people
seem to immediately write me off as some sort of religious lunatic or
think that I don't believe in AIs because I believe the brain necessarily
involves some sort of magic.

Um, no.

I don't believe in AIs because I know computers too well to believe that
they can think. I am, in other words, a skeptic. I don't believe it's
possible to construct a computer-based AI, and I invite the world to prove
me wrong. If the world does prove me wrong, I'm wrong, and will happily
change my opinion. My opinion on the subject is really not much different
than the opinion of a skeptic on any other subject.

My religious beliefs or lack thereof have absolutely nothing to do with
it.

> Yes, it's certainly possible that there's some "magic" involved (see
> above) that prevents it, but (a) it's unlikely, and (b) it's irrelevant:
> you asked me "why do people see AIs as attractive" not "what will we be
> able to do with AIs". I'm answering that question.

Understood. I took that as answered and moved on to discussing it; for
what it's worth, I grant you that there are reasons for wanting to
construct a computer-based AI.

>> And sure. You're right, I should amend my previous statement to
>> wondering why we'd care to make biological simulations or replicas of
>> humans.

> Where'd that come from, though?

At the time when I was writing it, I was thinking about the biological
replicas part of it. Sorry.

> Anyway you *can* see why people would consider them desirable, even if
> you don't see them as being likely. Which was the point.

Yup.

Peter da Silva

unread,

Apr 29, 1998, 3:00:00 AM4/29/98

to

In article <m3ra2hd...@windlord.Stanford.EDU>,

Russ Allbery <r...@stanford.edu> wrote:
>I do know what pissed me off enough to respond the way that I did. Every
>time I get started on AIs (yeah, this isn't a new topic with me), people
>seem to immediately write me off as some sort of religious lunatic or
>think that I don't believe in AIs because I believe the brain necessarily
>involves some sort of magic.

There's a reason for that. Honest. It's because you came out all in-yo-face
aggressive arguing that computers WILL NEVER be able to think because
thinking involves processes that can't be duplicated on a digital computer.

If that's how you usually respond to that speculation, then people are going
to categorize you based on that response. I'm not saying that's fair, but
it's an inevitable side effect of the expert systems that underly the
collective structures that we call "thinking" when we attempt to model
them in themselves.

>I don't believe in AIs because I know computers too well to believe that
>they can think.

I'm not exactly wet behind the ears when it comes to this newfangled
electric calculating thingummy either. I just don't think that thinking
is necessarily that big a deal. I'm not talking about computers that
have a rich internal life and hopes and dreams and aspirations (but I
rather suspect that's something we're going to have to work hard to
avoid once we get to the point of systems that can handle the rest of
the problems involved), but computers where you can sit down and talk
to them and get meaningful results out of the hash of broken metaphor
we call conversation, computers that can be left to manage for themselves
in real-life situations without inappropriate responses when faced with
problems that they weren't programmed for and their programmers didn't
anticipate in their design.

Which is quite a bit, really, when you consider how many humans don't.

But is that thinking?

Kai Henningsen

unread,

Apr 29, 1998, 3:00:00 AM4/29/98

to

pe...@taronga.com (Peter da Silva) wrote on 27.04.98 in <6i1s0c$t...@bonkers.taronga.com>:

> 1770. Lewis and Clark was enough of a Big Deal that many people simply

For some strange reason, I always first parse that as "Louis and Clark".

Kai Henningsen

unread,

Apr 29, 1998, 3:00:00 AM4/29/98

to

r...@stanford.edu (Russ Allbery) wrote on 28.04.98 in <m3ra2hd...@windlord.Stanford.EDU>:

> I don't believe in AIs because I know computers too well to believe that

> they can think. I am, in other words, a skeptic. I don't believe it's
> possible to construct a computer-based AI, and I invite the world to prove
> me wrong. If the world does prove me wrong, I'm wrong, and will happily
> change my opinion. My opinion on the subject is really not much different
> than the opinion of a skeptic on any other subject.

Well, I apply essentially the same argument to humans, with the result
that I can think of no good reason why AIs could be impossible.

And I'm sorry, but I can't say it any other way: I don't believe there's
any magic to thinking. But even though you deny it, that seems to be
exactly your point. At least I can see no other reason why you'd think AIs
to be impossible.

FWIW, I believe the essential human (or pre-human - I wasn't there)
breakthrough was not only doing a model of the universe, and putting
oneself in that model, but realizing there are lots of other themselves in
there as well, and what goes on in my brain is a good hint about what goes
on in theirs - thus including the model itself in the model. A multiply
recursive world view, if you want. Meta-abstraction. (Well, meta^n-
abstraction - once you get one meta, jumping to a lot more becomes easy.)

We can think about thinking.

Now, it was certainly hard to get to that point. But I don't think it'd be
particularly hard to put into a computer, once you have the rest going.
Design is infinitely easier than evolution.

Russ Allbery

unread,

Apr 29, 1998, 3:00:00 AM4/29/98

to

Kai Henningsen <kaih> writes:

> And I'm sorry, but I can't say it any other way: I don't believe there's
> any magic to thinking. But even though you deny it, that seems to be
> exactly your point. At least I can see no other reason why you'd think
> AIs to be impossible.

Depends on what definition you use for magic. I find Peter's definition
to be downright odd and am not using it. All I'm saying is that there's
no reason to expect it to be possible to map the analog human brain into a
digital program, and that I don't find the idea likely.

If you want to classify that belief as believing there's some "magic"
about thinking, that's fine; I don't intend to argue definitions. To me,
though, what "magic" implies is some divine breath of life or supernatural
intervention required for creation of intelligence, which is *not* what I
believe, and which is why I'm saying I don't believe there's magic
involved in thinking.

Russ Allbery

unread,

Apr 29, 1998, 3:00:00 AM4/29/98

to

Peter da Silva <pe...@taronga.com> writes:

> There's a reason for that. Honest. It's because you came out all
> in-yo-face aggressive arguing that computers WILL NEVER be able to think
> because thinking involves processes that can't be duplicated on a
> digital computer.

This is what happens when you only read half of a flamewar, Peter. I
wasn't in your face, I was in Lee's face. I apologize for coming across
as too adamant and pegging your buttons, but I'm human, and once in a
while I'm going to have a human reaction to twits.

> I'm not exactly wet behind the ears when it comes to this newfangled
> electric calculating thingummy either.

Nor was I implying that.

> I just don't think that thinking is necessarily that big a deal. I'm not
> talking about computers that have a rich internal life and hopes and
> dreams and aspirations (but I rather suspect that's something we're
> going to have to work hard to avoid once we get to the point of systems
> that can handle the rest of the problems involved), but computers where
> you can sit down and talk to them and get meaningful results out of the
> hash of broken metaphor we call conversation, computers that can be left
> to manage for themselves in real-life situations without inappropriate
> responses when faced with problems that they weren't programmed for and
> their programmers didn't anticipate in their design.

Mmm... see, this I can see a lot more easily. What you're colliding with
here in this discussion, and what I'm having a hard time communicating, is
that I have a hard time seeing computers doing things that I consider more
integral to intelligence than rational behavior in the face of unexpected
circumstances. Things like writing, fiction, artwork, philosophy, and the
like. Computers that have a culture. That's what I have a hard time
picturing computers doing.

*shrug* So maybe I'm wrong. It hardly matters if I am or not, since
*I'm* not working in that field and the people who are are going to
continue to do so regardless of whether I agree with them or not. But
until they actually produce computers that can do those sorts of things,
I'll remain skeptical.

It is, freely admitted, an argument from personal incredulity. I don't
expect to convince you that I'm right simply by saying I don't believe it.
I'm trying to point out to you how I *can* believe what I believe without
requiring supernatural cop-outs.

Russ Allbery

unread,

Apr 29, 1998, 3:00:00 AM4/29/98

to

Joel Baker <luc...@prophecy.lightbearer.com> writes:

> Ok. Question for the sake of discussion: what about analog hardware?

> (Please note that I'm not advocating any specific model here, or even
> claiming that we have a current model which could evolve this far,
> without some fundamental changes inthe model; I'm not familiar enough
> with the current edge of analog computing to speak to these points.)

I find that a lot more plausible. It won't surprise me at all if we're
eventually able to pretty much "construct" a human, including the mind,
using some technique or another. What I'm much less sure about is if the
better power system and in particular the ability to save and restore
global state will come along with that.

(I was thinking the other day about whether or not saving and restoring
global state in an intelligence is a survival characteristic. Hmm.)

> As a side note, dare I even mention the functioning (if trivial, yet)
> quantum computer that exists now? Or is that best left in the depths of
> the thread? :)

*laugh*

I haven't been following the field closely since I got out of grad school,
so I at least would be interested in references to the details.

Peter da Silva

unread,

Apr 29, 1998, 3:00:00 AM4/29/98

to

In article <m3zph46...@windlord.Stanford.EDU>,

Russ Allbery <r...@stanford.edu> wrote:
>This is what happens when you only read half of a flamewar, Peter.

Sorry, then. I suggest you do the same thing I did and just killfile Lee
so you're not tempted.

>Mmm... see, this I can see a lot more easily. What you're colliding with
>here in this discussion, and what I'm having a hard time communicating, is
>that I have a hard time seeing computers doing things that I consider more
>integral to intelligence than rational behavior in the face of unexpected
>circumstances. Things like writing, fiction, artwork, philosophy, and the
>like. Computers that have a culture. That's what I have a hard time
>picturing computers doing.

So you build a computer that gives you rational behaviour in the face of
unexpected circumstances, and you put it in a liberal arts university, and
you tell it that if it doesn't pass it'll be reprogrammed.

Perhaps the writing, fiction, artwork, and philosophy it comes up with are
the result of following rules and regurgitating past papers, but I'd bet
that a significant percentage of the HUMANS in the college are doing the
same thing. So are they not intelligent?

(yes, I'm assuming the thing passes, and there's enough of them doing this
that you can't get away with the imaginary student routine)

Peter da Silva

unread,

Apr 29, 1998, 3:00:00 AM4/29/98

to

In article <slrn6kffhk...@prophecy.lightbearer.com>,

Joel Baker <luc...@lightbearer.com> wrote:
>As a side note, dare I even mention the functioning (if trivial, yet)
>quantum computer that exists now? Or is that best left in the depths of
>the thread? :)

A hologram is a functioning (if trivial) quantum computer. It really is
the getting-past-trivial that's the hard part.

Joel Baker

unread,

Apr 30, 1998, 3:00:00 AM4/30/98

to

On 29 Apr 1998 17:01:44 -0700, Russ Allbery <r...@stanford.edu> wrote:

[ snip ]

> I find that a lot more plausible. It won't surprise me at all if we're
> eventually able to pretty much "construct" a human, including the mind,
> using some technique or another. What I'm much less sure about is if the
> better power system and in particular the ability to save and restore
> global state will come along with that.

It raises the question - how does one "save state" in a purely analog
system? Only another analog system is capable of it, and I *thought* that
most of those had problems with extended stability (though this could be
an entirely misguided notion.)

> (I was thinking the other day about whether or not saving and restoring
> global state in an intelligence is a survival characteristic. Hmm.)

Survival, yes. Wise... is much more questionable. The age-old sci-fi
fallback of "There's more born every year, but the old geezers refuse to
die anymore!"

[ snip ]

> *laugh*
>
> I haven't been following the field closely since I got out of grad school,
> so I at least would be interested in references to the details.

Found out about it today. A (none too specific or advanced) reference can
be found at: http://www1.sjmercury.com/scitech/center/quantum0428.htm
--
***************************************************************************
Joel Baker System Administrator - lightbearer.com
luc...@lightbearer.com http://www.lightbearer.com/~lucifer
The proof of the existence of truth is left as an exercise to the reader

Kai Henningsen

unread,

May 2, 1998, 3:00:00 AM5/2/98

to

r...@stanford.edu (Russ Allbery) wrote on 29.04.98 in <m3zph46...@windlord.Stanford.EDU>:

> Mmm... see, this I can see a lot more easily. What you're colliding with
> here in this discussion, and what I'm having a hard time communicating, is
> that I have a hard time seeing computers doing things that I consider more
> integral to intelligence than rational behavior in the face of unexpected
> circumstances. Things like writing, fiction, artwork, philosophy, and the
> like. Computers that have a culture. That's what I have a hard time
> picturing computers doing.

Then again, I happen to think that one is an almost unavoidable
consequence of the other. It just takes time (sometimes a lot).

Culture isn't what made us intelligent. It just happened because we became
intelligent.

And of course, it's possible to keep intelligent beings away from it. Has
been done with humans. Not that it's anything to be proud of.

I'm not sure how that will work out with artificial intelligences. But one
thing seems certain for me; we will have them *because* we *need* creative
computers.

And from what I hear, the creativity part is really not that hard,
compared to the useful understanding part. Essentially, and extremely over-
simplified, it amounts to putting some random number generators in.

> *shrug* So maybe I'm wrong. It hardly matters if I am or not, since
> *I'm* not working in that field and the people who are are going to
> continue to do so regardless of whether I agree with them or not. But
> until they actually produce computers that can do those sorts of things,
> I'll remain skeptical.

I think I already said that I estimate this to be a 50-200 year timescale.
If medicine doesn't do something pretty sensational, I have no real hope
of seeing it come true.

> It is, freely admitted, an argument from personal incredulity. I don't

Interestingly enough, that's exactly what my counter argument is. I guess
we just don't believe in different things :-)

> I'm trying to point out to you how I *can* believe what I believe without
> requiring supernatural cop-outs.

Hmm, well, I reserve the right to have my doubts about that.

Kai Henningsen

unread,

May 2, 1998, 3:00:00 AM5/2/98

to

pe...@taronga.com (Peter da Silva) wrote on 29.04.98 in <6i8vrn$4...@bonkers.taronga.com>:

> In article <slrn6kffhk...@prophecy.lightbearer.com>,
> Joel Baker <luc...@lightbearer.com> wrote:
> >As a side note, dare I even mention the functioning (if trivial, yet)
> >quantum computer that exists now? Or is that best left in the depths of
> >the thread? :)
>
> A hologram is a functioning (if trivial) quantum computer. It really is
> the getting-past-trivial that's the hard part.

A computer? Not for any definition of "computer" I consider useful. If a
hologram is a computer, then so's some mud at the bottom of a river bed -
it has memory and can recreate something similar to what created it.

But memory does not make a computer. Processing does. Holograms don't seem
to do that.

Russ Allbery

unread,

May 3, 1998, 3:00:00 AM5/3/98

to

Peter da Silva <pe...@taronga.com> writes:

> I suggest you do the same thing I did and just killfile Lee so you're
> not tempted.

Eh. I tend not to killfile people outside of news.admin.* if I can avoid
it; mostly just Grubor and Boursy. He's not that bad; I just need more
self control.

As long as I'm still responding to him and saying things I shouldn't
because he's posting, I need to not killfile him. Once I've managed to
ignore him myself, then I can killfile him in peace.

> So you build a computer that gives you rational behaviour in the face of
> unexpected circumstances, and you put it in a liberal arts university,
> and you tell it that if it doesn't pass it'll be reprogrammed.

Well, I think you can pass courses in a liberal arts university without
having an ounce of creativity or doing much more than recycling old
literary tricks. But then, I have a low opinion of courses on creativity
(and know people who strongly disagree with me on that). :)

> Perhaps the writing, fiction, artwork, and philosophy it comes up with
> are the result of following rules and regurgitating past papers, but I'd
> bet that a significant percentage of the HUMANS in the college are doing
> the same thing. So are they not intelligent?

I don't think they're exercising creativity very much in doing so. Hmm.
You're right, though, in that most of the arguments I'd present at this
point involve things you'd define as magic. (I wouldn't define them as
magic, though.)

But then, I have a different view of what goes into writing than a lot of
people do, and I realize that.

I'm losing the point of where I was going to go with this argument....

Oh! That's right. The factor that I think may be missing here is
emotion, since it's very difficult to interact with people without a very
firm grasp on human emotion. In particular, any sort of creativity tends
to strongly rely on emotion and the evoking of it.

Whether that's required for *intelligence* or not is an interesting
argument. (And I don't think emotion is immediately deriveable from
goals, logic, or self-preservation.)

Russ Allbery

unread,

May 3, 1998, 3:00:00 AM5/3/98

to

Kai Henningsen <kaih> writes:

> I'm not sure how that will work out with artificial intelligences. But
> one thing seems certain for me; we will have them *because* we *need*
> creative computers.

I definitely hear what you're saying there, and tend to agree with the
idea that *if* we end up getting intelligent computers, that's where
they're going to come from, not from some grand effort to build an AI.
(Which is not to say that I believe in emergent intelligence in computers
à la Terminator, or that I think that's the same thing.)

> And from what I hear, the creativity part is really not that hard,
> compared to the useful understanding part. Essentially, and extremely

> over-simplified, it amounts to putting some random number generators in.

I really, really dislike that model of creativity, but I do recognize that
most of my gut reaction to it is based in a notion of superiority ("I do
more than randomize things!") and therefore isn't valid logic.
Nonetheless, I think there's an intuitive aspect to creativity that's more
than just randomness.

> I think I already said that I estimate this to be a 50-200 year
> timescale. If medicine doesn't do something pretty sensational, I have
> no real hope of seeing it come true.

*nod* So chances are I'll just be skeptical for my life, and in the
meantime probably write sf about it. :)

>> I'm trying to point out to you how I *can* believe what I believe
>> without requiring supernatural cop-outs.

> Hmm, well, I reserve the right to have my doubts about that.

Fair enough. I certainly do believe in the supernatural, having some
personal experience with it (nope, can't prove it, no desire to try), and
I'll certainly agree that that affects my worldview and that it may be
subtly hidden in this line of reasoning deeply enough that I'm not seeing
it. But to sum up my basic belief, I think that digital computers don't
have enough capacity for intuitive or emergent behavior in large systems
to allow simulation of something that we would call intelligence.

(And there's nothing fundamentally supernatural about intuition in my
view; put at its most simplistic, it's derivation of knowledge via means
other than logical implication. Precisely the thing that computers have a
hard time with.)

Peter da Silva

unread,

May 3, 1998, 3:00:00 AM5/3/98

to

In article <m3n2czh...@windlord.Stanford.EDU>,

Russ Allbery <r...@stanford.edu> wrote:
>Oh! That's right. The factor that I think may be missing here is
>emotion, since it's very difficult to interact with people without a very
>firm grasp on human emotion. In particular, any sort of creativity tends
>to strongly rely on emotion and the evoking of it.

OK, let me present that argument again, without elaboration:

A lot of humans seem to get by without externally discernible amounts of
either. And if you can't tell from the outside is it a meaningful requirement
for something you're only going to be able to judge from the outside?

Kai Henningsen

unread,

May 3, 1998, 3:00:00 AM5/3/98

to

r...@stanford.edu (Russ Allbery) wrote on 03.05.98 in <m3k983h...@windlord.Stanford.EDU>:

> Kai Henningsen <kaih> writes:

> > And from what I hear, the creativity part is really not that hard,
> > compared to the useful understanding part. Essentially, and extremely
> > over-simplified, it amounts to putting some random number generators in.
>
> I really, really dislike that model of creativity, but I do recognize that
> most of my gut reaction to it is based in a notion of superiority ("I do
> more than randomize things!") and therefore isn't valid logic.

Of course you do. Unfiltered creativity is as useless as no creativity;
the value is created by filtering it. However, I see no reason why that
one wouldn't be part of the "useful understanding".

Or maybe put it like this: one thing is to come up with ideas for stories,
dialogues, etc. Another is to recognize which of these ideas is actually
good. Both together make for a good writer.

> Nonetheless, I think there's an intuitive aspect to creativity that's more
> than just randomness.

I strongly believe intuition is nothing more than pattern matching. One
important part of pattern matching, except for very primitive forms, is
that it is extremely hard to explain why something matches a pattern.

I seem to remember a very early experiment with letter recognition, where
someone built a large setup of common electronic components, and varied
their connections until some letter "seen" by an array of optical
receptors would produce a stronger response than any others.

Then, they tried to understand what that network did. They couldn't do it,
however.

That was probably a very early form of a trained neuronal network. And I
suspect the handling of this is very similar to what human brains do most
of the time - just tune the pattern recognition until it recognizes useful
things. Some are simple (hey, that's a traffic light), others are fairly
complicated (something in this room is not quite how it should be). The
latter gets called intuition, usually because we don't really understand
how our brain is coming to these conclusions.

Oh, and some are just plain wrong. "Hey, I can hear the sea in this
shell."

> > I think I already said that I estimate this to be a 50-200 year
> > timescale. If medicine doesn't do something pretty sensational, I have
> > no real hope of seeing it come true.
>
> *nod* So chances are I'll just be skeptical for my life, and in the
> meantime probably write sf about it. :)

I'm just afraid of the stage at the beginning of Hogan's book (I tend to
forget the English title because I read it in German, but I already
mentioned it somewhere recently - maybe here on nsm, maybe ng, maybe
nrf?).

Russ Allbery

unread,

Jun 8, 1998, 3:00:00 AM6/8/98

to

Following up to a *really* old article, so old that I doubt anyone cares
any more....

Peter da Silva <pe...@taronga.com> writes:

> Russ Allbery <r...@stanford.edu> wrote:

>> Oh! That's right. The factor that I think may be missing here is
>> emotion, since it's very difficult to interact with people without a
>> very firm grasp on human emotion. In particular, any sort of
>> creativity tends to strongly rely on emotion and the evoking of it.

> OK, let me present that argument again, without elaboration:

> A lot of humans seem to get by without externally discernible amounts of
> either. And if you can't tell from the outside is it a meaningful
> requirement for something you're only going to be able to judge from the
> outside?

This is a very good point.

Okay, I can accept that we may eventually see AI indiscernible from human
intelligence for those humans who do not demonstrate externally
discernible amounts of creativity. I'm still more skeptical about the
prospects of a creative computer.

Of course, so much of our *judgement* of creativity is tied up in
aesthetics that may not generalize to another type of intelligence, so
that skepticism may be essentially meaningless.

Russ Allbery

unread,

Jun 8, 1998, 3:00:00 AM6/8/98

to

Continuing to resurrect ancient articles (can you tell I'm catching up on
stuff I'd marked to come back to?)....

Kai Henningsen <kaih> writes:

> I strongly believe intuition is nothing more than pattern matching. One
> important part of pattern matching, except for very primitive forms, is
> that it is extremely hard to explain why something matches a pattern.

I honestly believe that there is more to intuition than that, but that
belief is based more on introspection and personal experience, which of
course could (a) be delusional on my part and (b) is unlikely to be
externally observable and therefore is in some sense irrelevant to the
question at hand. I have to concede this point, at least for this debate,
since I can't form a logical argument against it.

> I seem to remember a very early experiment with letter recognition,
> where someone built a large setup of common electronic components, and
> varied their connections until some letter "seen" by an array of optical
> receptors would produce a stronger response than any others.

> Then, they tried to understand what that network did. They couldn't do
> it, however.

> That was probably a very early form of a trained neuronal network.

Hasn't our understanding of exactly how neural networks do what they do
gotten much better since then? I remember it being something of a mystery
when they were first discovered, but now AI classes seem to teach the
theory behind them rather matter-of-factly.

> And I suspect the handling of this is very similar to what human brains
> do most of the time - just tune the pattern recognition until it
> recognizes useful things. Some are simple (hey, that's a traffic light),
> others are fairly complicated (something in this room is not quite how
> it should be). The latter gets called intuition, usually because we
> don't really understand how our brain is coming to these conclusions.

*nod*

What this question largely reduces to for me personally is what the
mechanism is whereby I "hear" character voices in my head, and where
writing at certain stages becomes largely a process of transcription for
my conscious mind.

It's that sort of event that I personally term "intuition"; I'm very
curious what the actual mechanics of it are.

Joel Baker

unread,

Jun 9, 1998, 3:00:00 AM6/9/98

to

On 08 Jun 1998 18:37:48 -0700, Russ Allbery <r...@stanford.edu> wrote:
> Continuing to resurrect ancient articles (can you tell I'm catching up on
> stuff I'd marked to come back to?)....
>
> Kai Henningsen <kaih> writes:
>
> > I strongly believe intuition is nothing more than pattern matching. One
> > important part of pattern matching, except for very primitive forms, is
> > that it is extremely hard to explain why something matches a pattern.
>
> I honestly believe that there is more to intuition than that, but that
> belief is based more on introspection and personal experience, which of
> course could (a) be delusional on my part and (b) is unlikely to be
> externally observable and therefore is in some sense irrelevant to the
> question at hand. I have to concede this point, at least for this debate,
> since I can't form a logical argument against it.
>
> > I seem to remember a very early experiment with letter recognition,
> > where someone built a large setup of common electronic components, and
> > varied their connections until some letter "seen" by an array of optical
> > receptors would produce a stronger response than any others.
>
> > Then, they tried to understand what that network did. They couldn't do
> > it, however.
>
> > That was probably a very early form of a trained neuronal network.
>
> Hasn't our understanding of exactly how neural networks do what they do
> gotten much better since then? I remember it being something of a mystery
> when they were first discovered, but now AI classes seem to teach the
> theory behind them rather matter-of-factly.

Advanced, yes, but I *think* that there is still a theory making the rounds
which has to do with being uncertain (in practical terms) of the state of
any system sufficiently complex enough to be useful. Something akin to doing
formal logic proofs of program correctness - possible, but essentially not
very useful when doing "real code" as opposed to Computer Science 301.

> > And I suspect the handling of this is very similar to what human brains
> > do most of the time - just tune the pattern recognition until it
> > recognizes useful things. Some are simple (hey, that's a traffic light),
> > others are fairly complicated (something in this room is not quite how
> > it should be). The latter gets called intuition, usually because we
> > don't really understand how our brain is coming to these conclusions.
>
> *nod*
>
> What this question largely reduces to for me personally is what the
> mechanism is whereby I "hear" character voices in my head, and where
> writing at certain stages becomes largely a process of transcription for
> my conscious mind.

Notion: partial pattern matching can "fill in" missing pieces, given a
framework of context, and a selection of choices, similar to doing the
missing bits of a jigsaw puzzle. Is it possible to conceive of a situation
in which the voices are based on taking a partial match against a set
of archetypes (both plot and character based) and picking semi-random
choices from what "makes sense" given that context? Once the core of the
pattern is selected, filling in the rest is, indeed, largely a matter
which could be done on the fly, perhaps.

Or perhaps not. :)

> It's that sort of event that I personally term "intuition"; I'm very
> curious what the actual mechanics of it are.

--

Kai Henningsen

unread,

Jun 9, 1998, 3:00:00 AM6/9/98

to

r...@stanford.edu (Russ Allbery) wrote on 08.06.98 in <m37m2rw...@windlord.Stanford.EDU>:

> Continuing to resurrect ancient articles (can you tell I'm catching up on
> stuff I'd marked to come back to?)....

I don't even remember most of the context, but let's see ...

> Kai Henningsen <kaih> writes:

> > Then, they tried to understand what that network did. They couldn't do
> > it, however.

> Hasn't our understanding of exactly how neural networks do what they do

> gotten much better since then? I remember it being something of a mystery
> when they were first discovered, but now AI classes seem to teach the
> theory behind them rather matter-of-factly.

Oh, I have no doubt that the mechanics are well understood. However, that
doesn't mean that you can look at the complete internal state of such a
network and explain just what it recognizes in the data it is fed. It
doesn't even mean that, if you *know* what it recognizes, you can explain
*why* it recognizes this.

Except for very simple networks, that is.

Maybe compare it to some piece of art, and trying to explain why it moves
you.

A better analogy (though useful to fewer people) is probably to explain
what a given part of DNA will do. We do know the relevant biochemistry
principles, but that doesn't help very much ...

> What this question largely reduces to for me personally is what the
> mechanism is whereby I "hear" character voices in my head, and where
> writing at certain stages becomes largely a process of transcription for
> my conscious mind.
>

> It's that sort of event that I personally term "intuition"; I'm very
> curious what the actual mechanics of it are.

Well, you'll probably not be very happy with my theory of random generator
+ memories + picking out the good parts (where recognizing what's good is
another pattern matcher) ... works for me, though. I'm continually amazed
what my brain can come up with (especially in dreams), and I've given up
writing because I've found that I am much better at picking out what
doesn't work on paper, than finding something that does. In a dream, the
suspension of disbelief is more-or-less built in - and still doesn't work
all the time.

Kai Henningsen

unread,

Jun 9, 1998, 3:00:00 AM6/9/98

to

r...@stanford.edu (Russ Allbery) wrote on 08.06.98 in <m3af7nw...@windlord.Stanford.EDU>:

> I'm still more skeptical about the
> prospects of a creative computer.
>
> Of course, so much of our *judgement* of creativity is tied up in
> aesthetics that may not generalize to another type of intelligence, so
> that skepticism may be essentially meaningless.

I'm not so sure about that. There's a lot of it outside the arts.

Have you read "The two faces of tomorrow"? The problem it starts out with
is a computer network that is creative all right, it's just not
intelligent enough to recognize all the constraints. Like, to use an
example not in the book, want to get rid of the fleas on your dog? Just
heat dog to 300 degrees Celsius. All fleas will be dead. Oops, you didn't
tell me the dog was supposed to survive this ...

And it doesn't seem so far away from what we _know_ computers can do. Just
combine the facts you know about until you get a result that seems to do
what was desired. Pretty mechanical, all you need (besides some way to
represent facts) is a lot of raw processing power.

I don't think creativity and intelligence are closely related. (Not
completely unrelated, either.)

Russ Allbery

unread,

Jun 21, 1998, 3:00:00 AM6/21/98

to

Kai Henningsen <kaih> writes:
> r...@stanford.edu (Russ Allbery) writes:

>> Hasn't our understanding of exactly how neural networks do what they do
>> gotten much better since then? I remember it being something of a
>> mystery when they were first discovered, but now AI classes seem to
>> teach the theory behind them rather matter-of-factly.

> Oh, I have no doubt that the mechanics are well understood. However,
> that doesn't mean that you can look at the complete internal state of
> such a network and explain just what it recognizes in the data it is
> fed.

That's okay, I can't look at assembly code or badly written BASIC programs
and explain what they're supposed to do either, at least easily. I'd
expect that with enough work one could deconstruct the effect of a neural
network. I think that's most of what I'm asking. If one *can't*, even
with the amount of work it takes to disassemble binary code, that would be
very interesting.

> It doesn't even mean that, if you *know* what it recognizes, you can
> explain *why* it recognizes this.

I'm not quite sure what that means. In other words, what's the difference
between being able to tell what a neural network recognizes and why it
recognizes that? (It recognizes it because of the relative weights on the
neurons and the connection grid....)

>> What this question largely reduces to for me personally is what the
>> mechanism is whereby I "hear" character voices in my head, and where
>> writing at certain stages becomes largely a process of transcription
>> for my conscious mind.

>> It's that sort of event that I personally term "intuition"; I'm very
>> curious what the actual mechanics of it are.

> Well, you'll probably not be very happy with my theory of random
> generator + memories + picking out the good parts (where recognizing
> what's good is another pattern matcher) ... works for me, though. I'm
> continually amazed what my brain can come up with (especially in
> dreams), and I've given up writing because I've found that I am much
> better at picking out what doesn't work on paper, than finding something
> that does. In a dream, the suspension of disbelief is more-or-less built
> in - and still doesn't work all the time.

I guess my dreams/writing (the distinction is rather blurred for me) have
never felt like random generation plus memories. Of course, we're still
in the "feel like" realm of discussion, which makes nigh everything I say
impossible to defend with logic.

Russ Allbery

unread,

Jun 21, 1998, 3:00:00 AM6/21/98

to

Kai Henningsen <kaih> writes:
> r...@stanford.edu (Russ Allbery) writes:

>> I'm still more skeptical about the prospects of a creative computer.

>> Of course, so much of our *judgement* of creativity is tied up in
>> aesthetics that may not generalize to another type of intelligence, so
>> that skepticism may be essentially meaningless.

> I'm not so sure about that. There's a lot of it outside the arts.

I think we're probably using different definitions of creativity. I tend
to have that problem, given that two of the characters I write are musae.
It focuses my thinking on art a lot more, rather than encompassing things
like "creative solutions to problems." (As someone who is both a writer
and a programmer, I think the two are very different types of creativity;
they certainly are sparked by totally different things for me, occur at
very different ends of my emotional cycle, and don't involve the same
thinking patterns at all.)

> Have you read "The two faces of tomorrow"? The problem it starts out
> with is a computer network that is creative all right, it's just not
> intelligent enough to recognize all the constraints. Like, to use an
> example not in the book, want to get rid of the fleas on your dog? Just
> heat dog to 300 degrees Celsius. All fleas will be dead. Oops, you
> didn't tell me the dog was supposed to survive this ...

Haven't read it, but that's very plausible to me. That sounds like an
expert system reasoning based on incomplete knowledge and rulesets from
one of those "common sense" knowledge bases.

> And it doesn't seem so far away from what we _know_ computers can
> do. Just combine the facts you know about until you get a result that
> seems to do what was desired. Pretty mechanical, all you need (besides
> some way to represent facts) is a lot of raw processing power.

And a huge database. This is sounds like the "teach computers common
sense" project. But to me, getting rid of fleas by heating the dog is a
totally different kind of creativity than writing a story, so much so that
they deserve two separate words.

Kai Henningsen

unread,

Jun 21, 1998, 3:00:00 AM6/21/98

to

r...@stanford.edu (Russ Allbery) wrote on 21.06.98 in <m3u35ex...@windlord.Stanford.EDU>:

> Kai Henningsen <kaih> writes:
> > r...@stanford.edu (Russ Allbery) writes:
>

> >> Hasn't our understanding of exactly how neural networks do what they do
> >> gotten much better since then? I remember it being something of a
> >> mystery when they were first discovered, but now AI classes seem to
> >> teach the theory behind them rather matter-of-factly.
>
> > Oh, I have no doubt that the mechanics are well understood. However,
> > that doesn't mean that you can look at the complete internal state of
> > such a network and explain just what it recognizes in the data it is
> > fed.
>
> That's okay, I can't look at assembly code or badly written BASIC programs
> and explain what they're supposed to do either, at least easily. I'd
> expect that with enough work one could deconstruct the effect of a neural
> network. I think that's most of what I'm asking. If one *can't*, even
> with the amount of work it takes to disassemble binary code, that would be
> very interesting.

Well, any problem is solvable when you throw enough resources at it -
however, it's not clear that the human mind has enough resources to
understand a (concrete) complex neuronal net.

Also, I suspect this problem may have similarities to asking "just why is
the third digit in pi a four?" - is there really a better answer than
"that just happens to be the way it is"?

> > It doesn't even mean that, if you *know* what it recognizes, you can
> > explain *why* it recognizes this.
>
> I'm not quite sure what that means. In other words, what's the difference
> between being able to tell what a neural network recognizes and why it
> recognizes that? (It recognizes it because of the relative weights on the
> neurons and the connection grid....)

Well, you can see what it recognizes just by observing it. And you can
observe the relative weights and the connection grid.

Does that necessarily mean you can explain just why the combination of
those weights and connections leads to recognizing these things? And, is
there a way to predict what a change in weights and/or connections will do
to the set of recognized things without actually trying it out?

Remember chaos theory. Remember the three body problem. We understand
everything that goes into the problem, yet the only way to find a solution
is still to play it through and observe what happens. And that one is
rather a lot simpler than a complex neuronal network.

> > Well, you'll probably not be very happy with my theory of random
> > generator + memories + picking out the good parts (where recognizing
> > what's good is another pattern matcher) ... works for me, though. I'm
> > continually amazed what my brain can come up with (especially in
> > dreams), and I've given up writing because I've found that I am much
> > better at picking out what doesn't work on paper, than finding something
> > that does. In a dream, the suspension of disbelief is more-or-less built
> > in - and still doesn't work all the time.
>
> I guess my dreams/writing (the distinction is rather blurred for me) have
> never felt like random generation plus memories. Of course, we're still
> in the "feel like" realm of discussion, which makes nigh everything I say
> impossible to defend with logic.

Aah, but that's why I said suspension of disbelief is built into dreams.
Of course they don't *feel* like that. (And they _still_ sometimes feel
pretty made-up :-))

However, I haven't found anything in them that wouldn't be possible to
interpret that way. Which satisfies my sense of aesthetics: as a
programmer, I'm all for creation of complex effects by combining very few
basic algorithms.

Kai Henningsen

unread,

Jun 21, 1998, 3:00:00 AM6/21/98

to

r...@stanford.edu (Russ Allbery) wrote on 21.06.98 in <m3ra0ix...@windlord.Stanford.EDU>:

> Kai Henningsen <kaih> writes:
> > r...@stanford.edu (Russ Allbery) writes:
>

> >> I'm still more skeptical about the prospects of a creative computer.
>
> >> Of course, so much of our *judgement* of creativity is tied up in
> >> aesthetics that may not generalize to another type of intelligence, so
> >> that skepticism may be essentially meaningless.
>
> > I'm not so sure about that. There's a lot of it outside the arts.
>
> I think we're probably using different definitions of creativity. I tend

Maybe. I think of programming as a creative occupation, for example, just
like writing or painting, in that an essential part of it is coming up
with new ideas - and like any creative occupation, one needs to sort out
which of these new ideas are actually worth pursuing, and which are just
crap.

> to have that problem, given that two of the characters I write are musae.
> It focuses my thinking on art a lot more, rather than encompassing things
> like "creative solutions to problems." (As someone who is both a writer
> and a programmer, I think the two are very different types of creativity;
> they certainly are sparked by totally different things for me, occur at
> very different ends of my emotional cycle, and don't involve the same
> thinking patterns at all.)

That doesn't seem to be true for me. I won't call myself an artist, but I
have done a (very) few things in those areas - it's practically impossible
to avoid - and I think it was actually very similar to programming. Well,
to the creative part of programming, which for me often (not always)
happens far away from any computer.

> > Have you read "The two faces of tomorrow"? The problem it starts out
> > with is a computer network that is creative all right, it's just not
> > intelligent enough to recognize all the constraints. Like, to use an
> > example not in the book, want to get rid of the fleas on your dog? Just
> > heat dog to 300 degrees Celsius. All fleas will be dead. Oops, you
> > didn't tell me the dog was supposed to survive this ...
>
> Haven't read it, but that's very plausible to me. That sounds like an
> expert system reasoning based on incomplete knowledge and rulesets from
> one of those "common sense" knowledge bases.
>
> > And it doesn't seem so far away from what we _know_ computers can
> > do. Just combine the facts you know about until you get a result that
> > seems to do what was desired. Pretty mechanical, all you need (besides
> > some way to represent facts) is a lot of raw processing power.
>
> And a huge database. This is sounds like the "teach computers common
> sense" project. But to me, getting rid of fleas by heating the dog is a
> totally different kind of creativity than writing a story, so much so that
> they deserve two separate words.

Well, it's different from writing a _good_ story, just as dog heating is
different from a _good_ way to get rid of fleas.

But I won't be astonished if someone came up with a program to write stuff
like those awful hack SF stories often mentioned on Usenet. What was it -
eye of aragon or something like that? Formula + random numbers sure ought
to be good enough for it.

And, of course, the human mind _has_ both a huge database, and a lot of
raw processing power.

(As to the book, the "teach common sense" part was what it was all about.)

Russ Allbery

unread,

Jun 22, 1998, 3:00:00 AM6/22/98

to

As an aside, does anyone know the correct blunt instrument and sensitive
point that should be introduced to prevent SuperCite from chopping off
e-mail addresses at the first character it doesn't think it understands?

Kai Henningsen <kaih> writes:

> Also, I suspect this problem may have similarities to asking "just why
> is the third digit in pi a four?" - is there really a better answer than
> "that just happens to be the way it is"?

*nod* Agreed, and I think that's a little bit of what I was trying to get
at later on. Hm. I guess the general idea that I'm taking out of this is
that I don't find creativity a particular procedural activity, and
therefore for the most part computers don't seem creative as they're
largely procedural, but aspects of computers that *aren't* procedural
(like neural nets) stand a better chance of approaching that kind of
functionality.

> Does that necessarily mean you can explain just why the combination of
> those weights and connections leads to recognizing these things? And, is
> there a way to predict what a change in weights and/or connections will
> do to the set of recognized things without actually trying it out?

Good point. This is something I'd be interested in knowing.

> Remember chaos theory. Remember the three body problem. We understand
> everything that goes into the problem, yet the only way to find a
> solution is still to play it through and observe what happens. And that
> one is rather a lot simpler than a complex neuronal network.

Well... I suppose. It depends on definitions of complexity, I think. But
yeah. (Much of human thinking has always struck me as being very
susceptible to the butterfly effect.)

> Aah, but that's why I said suspension of disbelief is built into dreams.
> Of course they don't *feel* like that. (And they _still_ sometimes feel
> pretty made-up :-))

> However, I haven't found anything in them that wouldn't be possible to
> interpret that way. Which satisfies my sense of aesthetics: as a
> programmer, I'm all for creation of complex effects by combining very
> few basic algorithms.

I think my "problem," if one wants to consider it that, is that my
disbelief is pretty thoroughly suspended in a few places. I think that's
a good thing, as it lets me write without thinking overly much about *how*
I write, which in my case leads to nothing but a descending spiral of
pointless self-examination.

Russ Allbery

unread,

Jun 22, 1998, 3:00:00 AM6/22/98

to

Kai Henningsen <kaih> writes:

> Well, it's different from writing a _good_ story, just as dog heating is
> different from a _good_ way to get rid of fleas.

> But I won't be astonished if someone came up with a program to write
> stuff like those awful hack SF stories often mentioned on Usenet. What
> was it - eye of aragon or something like that? Formula + random numbers
> sure ought to be good enough for it.

> And, of course, the human mind _has_ both a huge database, and a lot of
> raw processing power.

This is all true. At the same time, I write via going into a trance state
(see current net.writing.general traffic) and reaching the point where I'm
watching characters act. Which, I realize, can't be shown to be more than
random generation plus good pattern recognition, but which mentally bears
little resemblence to that (and has *nothing* to do with the way I
program; I disconnect logic almost entirely when writing).

It's one of those questions that keeps making me want to ask somewhat
stupid questions like whether or not writing would feel the same way to a
computer as it does to me.

Kai Henningsen

unread,

Jun 29, 1998, 3:00:00 AM6/29/98

to

r...@stanford.edu (Russ Allbery) wrote on 22.06.98 in <m3iulto...@windlord.Stanford.EDU>:

> This is all true. At the same time, I write via going into a trance state
> (see current net.writing.general traffic) and reaching the point where I'm
> watching characters act. Which, I realize, can't be shown to be more than
> random generation plus good pattern recognition, but which mentally bears
> little resemblence to that (and has *nothing* to do with the way I
> program; I disconnect logic almost entirely when writing).

Well, I can't say much about writing.

As to programming - hmmm ... I won't call it a trance state, but there is
often quite a lot going on on the subconcious level. Quite a number of the
decisions I make come from some sort of "feeling" about the problem at
hand. I often "just know" where I should create another abstraction, for
example.

OTOH, explicit reasoning also plays a significant part - "suppose I do it
like that, can I handle this? - and this? - oops, that one breaks, try
another solution".

On the gripping hand, I do use aesthetics as a basis for decisions. "This
solution works, but it is so ugly I need to rewrite it." (Or even "so I
won't do it like that, but try to find a cleaner solution instead".)

> It's one of those questions that keeps making me want to ask somewhat
> stupid questions like whether or not writing would feel the same way to a
> computer as it does to me.

You'll have to wait until intelligent computers can read&write Usenet for
an answer to that one :-)

(Scary thought - computer like that going through DejaNews ...)

Russ Allbery

unread,

Jul 1, 1998, 3:00:00 AM7/1/98

to

Kai Henningsen <kaih> writes:

> Well, I can't say much about writing.

> As to programming - hmmm ... I won't call it a trance state, but there
> is often quite a lot going on on the subconcious level. Quite a number
> of the decisions I make come from some sort of "feeling" about the
> problem at hand. I often "just know" where I should create another
> abstraction, for example.

Oh, yeah, me too, but I definitely wouldn't call it a trance state. My
logical mind is very much engaged when I'm programming, whereas I would
describe the process I go through when writing as turning it off.

> OTOH, explicit reasoning also plays a significant part - "suppose I do
> it like that, can I handle this? - and this? - oops, that one breaks,
> try another solution".

*nod*

> On the gripping hand, I do use aesthetics as a basis for
> decisions. "This solution works, but it is so ugly I need to rewrite
> it." (Or even "so I won't do it like that, but try to find a cleaner
> solution instead".)

*nod* Me too. But for me it's a different sense of aethetics.

> You'll have to wait until intelligent computers can read&write Usenet
> for an answer to that one :-)

Yup!

> (Scary thought - computer like that going through DejaNews ...)

Heh. I protect myself against things like that by posting ungodly numbers
of articles! :)

Kai Henningsen

unread,

Jul 7, 1998, 3:00:00 AM7/7/98

to

g...@hugo.westfalen.de (Georg Bauer) wrote on 04.07.98 in <gb-040798...@hugo.westfalen.de>:

> becomes directly writing my thoughts. I often have the feeling that I
> think in a programming language instead thinking and reasoning about a
> program - it helps to use the language that's needed to communicate your

Hah.

On times, I've been dreaming in a weird sort of "language" that consisted
only of control flow decisions. No subject matter or other context, just
doing one control flow decision after another in hopes of reaching some
unspecified goal. Truely weird.

rif...@cybernothing.org

unread,

Jul 8, 1998, 3:00:00 AM7/8/98

to

Russ Allbery <r...@stanford.edu> wrote:
>Oh, yeah, me too, but I definitely wouldn't call it a trance state. My
>logical mind is very much engaged when I'm programming, whereas I would

Anyone else read that as "enraged when I'm programming"?

Maybe it's just me...

rif...@afn.org : Let me take you down (without a sound),
Jeff The Riffer : Dead before you hit the ground.
Drifter... : Blood washes my hand (can't understand),
Homo Postmortemus : Sterlizing my pain.

Steve Watt

unread,

Jul 8, 1998, 3:00:00 AM7/8/98

to

Russ Allbery <r...@stanford.edu> wrote:
>Kai Henningsen <kaih> writes:
>
>> Well, I can't say much about writing.
>
>> As to programming - hmmm ... I won't call it a trance state, but there
>> is often quite a lot going on on the subconcious level. Quite a number
>> of the decisions I make come from some sort of "feeling" about the
>> problem at hand. I often "just know" where I should create another
>> abstraction, for example.
>

>Oh, yeah, me too, but I definitely wouldn't call it a trance state. My
>logical mind is very much engaged when I'm programming, whereas I would

>describe the process I go through when writing as turning it off.

Wow. I think you just hit my nail on the head. I've always been
told that I'm a good writer (either fiction, or spec/design/user
docs), but I find it ... painful is the only word I can locate ...
to do. Fiction is *soooo* much easier to write, because it's the
trance-like flow state that doesn't involve the logical mind -- I
just write the scene as it happens. Fortunately, I'm generally
literate enough to make good word choices while in such a flow.

But tech docs? I have to turn my "writer" on and off, and then my
"geek." I can't explain it any more than I am able to explain the
Dvorak/QWERTY switch that I have in my head. And if you want to
mess me up, put me in front of a computer other than my default
workstations that are Dvorak. I've been told that I'm entertaining
to watch while I throw the mental switch.

>> On the gripping hand, I do use aesthetics as a basis for

[ OK, where's this ref from? I remember it, but don't remember
where I saw it. ]

>> decisions. "This solution works, but it is so ugly I need to rewrite
>> it." (Or even "so I won't do it like that, but try to find a cleaner
>> solution instead".)
>
>*nod* Me too. But for me it's a different sense of aethetics.

Actually, I have found that the aesthetics are important, from a cost
standpoint; if a particular solution is not aesthetically pleasing,
it is (in my experience) much more expensive once in maintenance.
When it's clean, it can be picked up easily.

Which rules out 99% of *my* Perl, but almost none of my C. ;)

>Heh. I protect myself against things like that by posting ungodly numbers
>of articles! :)

And yet you manage to make all of them well-reasoned. Wish I had enough
time to debate Dave Hayes and do other things as well. I've even
given up on debating rfg, but that could just be because he's such
a jerk.
--
Steve Watt KD6GGD PP-ASEL-IA Packet: KD6GGD @ N0ARY.#NOCAL.CA.USA.NA
ICBM: 121W 56' 58.1" / 37N 20' 14.2" Internet: steve @ Watt.COM
Free time? There's no such thing. It just comes in varying prices...

Lars Balker Rasmussen

unread,

Jul 8, 1998, 3:00:00 AM7/8/98

to

kaih=6xQSR...@khms.westfalen.de (Kai Henningsen) writes:
> On times, I've been dreaming in a weird sort of "language" that consisted
> only of control flow decisions. No subject matter or other context, just
> doing one control flow decision after another in hopes of reaching some
> unspecified goal. Truely weird.

Ah, you've dreamt Tetris too?
--
Lars Balker Rasmussen "Woo hoo!?"

Kai Henningsen

unread,

Jul 9, 1998, 3:00:00 AM7/9/98

to

st...@Watt.COM (Steve Watt) wrote on 08.07.98 in <EvrF7...@Watt.COM>:

> Russ Allbery <r...@stanford.edu> wrote:

> >Oh, yeah, me too, but I definitely wouldn't call it a trance state. My
> >logical mind is very much engaged when I'm programming, whereas I would
> >describe the process I go through when writing as turning it off.
>
> Wow. I think you just hit my nail on the head. I've always been
> told that I'm a good writer (either fiction, or spec/design/user
> docs), but I find it ... painful is the only word I can locate ...
> to do. Fiction is *soooo* much easier to write, because it's the
> trance-like flow state that doesn't involve the logical mind -- I
> just write the scene as it happens. Fortunately, I'm generally
> literate enough to make good word choices while in such a flow.

Somehow, this is frightening.

You see, _reading_ (yes, even reading fiction) is for me an activity that
has my logical mind very much active. (And makes me curse when the author
does something particularly illogical.)

Shutting off that part of my mind ... I think that's what they call
"asleep". That is, I don't think I _can_ shut it off while awake.

> >> On the gripping hand, I do use aesthetics as a basis for
>
> [ OK, where's this ref from? I remember it, but don't remember
> where I saw it. ]

The Motie books.

Kai Henningsen

unread,

Jul 9, 1998, 3:00:00 AM7/9/98

to

gn...@daimi.aau.dk (Lars Balker Rasmussen) wrote on 08.07.98 in <0fbtr0k...@humulus.daimi.aau.dk>:

Nope. I don't like games with time-limit stress. Besides, there's *no*
geometric component to those dreams.

r...@greenend.org.uk

unread,

Jul 9, 1998, 3:00:00 AM7/9/98

to

Lars Balker Rasmussen <gn...@daimi.aau.dk> writes:
> kaih=6xQSR...@khms.westfalen.de (Kai Henningsen) writes:

>> On times, I've been dreaming in a weird sort of "language" that
>> consisted only of control flow decisions. No subject matter or
>> other context, just doing one control flow decision after another
>> in hopes of reaching some unspecified goal. Truely weird.
>
> Ah, you've dreamt Tetris too?

Never Tetris myself, but I do sometimes see images from various other
games - Warcrafy and Starcraft in particular - in front of my eyes,
often when awake but with my eyes closed. Other players of the same
games have reported similar effects. It's straneg...

(Hmm, I'm going to come across as totally obsessed with these games on
net.*/bofh.* at this. Oh well...)

Russ Allbery

unread,

Jul 10, 1998, 3:00:00 AM7/10/98

to

Note crosspost.

Kai Henningsen <kaih> writes:

> Somehow, this is frightening.

> You see, _reading_ (yes, even reading fiction) is for me an activity
> that has my logical mind very much active. (And makes me curse when the
> author does something particularly illogical.)

Oh, yes, mine too. No doubt about it. I stay very much logically
involved in fiction that I read, and I'm analyzing the logic of it as I
read it. Unless it's *really* intense, in which case sometimes that shuts
down for a while.

Writing is totally different.

With writing, I'm trying to turn off as much of me as possible to let the
character come to the fore, which includes my logical mind. And this is
particularly true of the way that I write, since most of my "hooks" (the
things I concentrate to get in frame for the character) are all emotional
in nature. I can feel what the character is feeling, and that's what lets
me go the rest of the way into letting them dictate through my hands into
the story/scene/whatever.

(I'm often laughing, crying, angry, or otherwise openly emotional while
I'm writing; that's one of the reasons why I strongly prefer to be
physically alone while doing it.)

> Shutting off that part of my mind ... I think that's what they call
> "asleep". That is, I don't think I _can_ shut it off while awake.

It's one of the first things they try to teach you in meditation, or so
I've been told. I've had people try to explain to me how to do it,
according to the "standard" ways, and none of them work for me at all, but
concentrating on a character will do it almost immediately.

Russ Allbery

unread,

Jul 10, 1998, 3:00:00 AM7/10/98

to

Note crosspost; I think it's appropriate.

Steve Watt <st...@Watt.COM> writes:
> Russ Allbery <r...@stanford.edu> wrote:
>> Kai Henningsen <kaih> writes:

>>> decisions. "This solution works, but it is so ugly I need to rewrite
>>> it." (Or even "so I won't do it like that, but try to find a cleaner
>>> solution instead".)

>> *nod* Me too. But for me it's a different sense of aethetics.

> Actually, I have found that the aesthetics are important, from a cost
> standpoint; if a particular solution is not aesthetically pleasing, it
> is (in my experience) much more expensive once in maintenance. When
> it's clean, it can be picked up easily.

> Which rules out 99% of *my* Perl, but almost none of my C. ;)

I definitely agree that aesthetics are important, and they have a lot to
do with how I write code. That was what started the whole argument in
net.computers.other about code style and the like. :) My aesthetics for
code are just very different than my aesthetics for writing.

I *can* apply my code aesthetics to writing. That's how I design worlds;
that uses some of the same circuits. But when it comes to characters,
it's a completely different thing. I judge characters by how much they
want to be written, how much they talk to me, how easily they come in
trace state, that sort of thing. None of which feels anything like how I
write, at least to me.

>> Heh. I protect myself against things like that by posting ungodly
>> numbers of articles! :)

> And yet you manage to make all of them well-reasoned.

Heh. Not all of them. :)

> Wish I had enough time to debate Dave Hayes and do other things as well.

If you think the newsgroup postings are bad, you should see the e-mail. I
know I've gone over 600 lines before.

> I've even given up on debating rfg, but that could just be because he's
> such a jerk.

I still get along with him, but I don't bother debating him any more.

Georg Bauer

unread,

Jul 12, 1998, 3:00:00 AM7/12/98

to

In article <wwvww9m...@sfere.greenend.org.uk>, r...@greenend.org.uk wrote:

>games - Warcrafy and Starcraft in particular - in front of my eyes,
>often when awake but with my eyes closed. Other players of the same
>games have reported similar effects. It's straneg...

Hmm. Makes me wonder if there is something like a screensaver for your
eyes to prevent the burn-in - because I think that is what may happened to
you during you long hours playing those games ;-)

bye, Georg

--
"Sicher ist, das nichts sicher ist. Aber selbst das nicht."
- Ringelnatz

Josh Brandt

unread,

Jul 14, 1998, 3:00:00 AM7/14/98

to

In article <gb-120798...@hugo.westfalen.de>,

Georg Bauer <g...@hugo.westfalen.de> wrote:
>Hmm. Makes me wonder if there is something like a screensaver for your
>eyes to prevent the burn-in - because I think that is what may happened to
>you during you long hours playing those games ;-)

I've dreamed Marathon. Including the level editor.

Josh
"That was a cool dream. I wonder how it worked... Let's edit it!"

--
...said it was heaven just to breathe your air Severed Heads
J. Brandt - mu...@sidehack.gweep.net

Bruce Baugh

unread,

Aug 1, 1998, 3:00:00 AM8/1/98

to

In article <6xWVX...@khms.westfalen.de>, Kai Henningsen wrote:

> You see, _reading_ (yes, even reading fiction) is for me an activity that
> has my logical mind very much active. (And makes me curse when the author
> does something particularly illogical.)

> Shutting off that part of my mind ... I think that's what they call
> "asleep". That is, I don't think I _can_ shut it off while awake.

Well, I dabble in extremely high-level programming languages and hang
around vastly more competent programmers, and make my living as a
freelance writer. What I do in writing sounds a lot like what my
friends do when programming - my whole mind is engaged, with logical
analysis of my goals, research, and so on fused with intuitive leaps,
semi-consciously remembered snipped of previous things to include, and
like that.

I recall reading some years ago that CAT scans showed that while
non-musicians use only certain specific portions of the brain when
listening to music, musicians use a lot more, with centers dedicated
to physical performance and rational thought coming into play as well
as aesthetic interpretation. I would be unsurprised to find that the
same is true of both writing and programming.

--
Bruce Baugh
bruce...@mindspring.com
http://brucebaugh.home.mindspring.com/ - New sf by
S.M. Stirling, rolegaming, writers' tools