Danny Kodicek wrote:

So I think it is quite an empty assertion to make at this time that you
can simulate some ultra complex process that we hardly understand. And
what about the possibility that there is noise in the brain, or chaotic
processes going on, or the randomness due to quantum effects?

To deal with the last point first, randomness and chaos are almost

a red herring (after all, would it make you more comfortable to feel

your emotions are based on chaos than on an algorithm? Does that make

somehow more 'free'?). Besides which, there's no reason a computer

can't use randomness (many AI programs do), and if it makes you any

for them to get that randomness from a quantum source rather than a
pseudo-random algorithm, there's absolutely nothing to stop them doing

I really can't see how that could make an effective difference to the

result.

The argument in the thread is about simulating the human brain with a
digital computer or network of such computers. If such a simulation is
to be successful you should get similar outputs from it given a set of
inputs as you would from the brain. If you have randomness built into
both the simulation and the brain, this will not happen.

Consider two programs which partly rely on pseudorandom number
generation -- genetic algorithm programs for example. If you start both
programs going with different seed numbers, they will produce different
results. I've written GA programs which produce different results each
time they are run because they use the operating system timestamp as a
random number generator seed.

As for programming emotions, no one would 'program a computer to be

Happiness is far too complex a state to be dealt with in that way.

are a mix of any number of different processes (including, in humans,
various hormonal and pheromonal effects, as Andrew mentions). In a

AI program, the same would be the case.

You're talking about AI here, not accurately simulating the human brain
(or mind). I'm not objecting to AI. What I'm objecting to is the
notion that the mind itself can be simulated on a computer powerful
enough.

To perform such a simulation

aspects of consciousness from it, including every possible gradation of
emotion.

AI is fine as far as it goes. It may do things completely differently
than minds do and still obtain useful results. Chess playing
programs/computers are an example. Gary Kasparov doesn't play chess as
well as he does by searching through 200 million possible moves every
second like Deep Blue does.

I replied to an earlier post of yours but for some reason it didn't show

(I think my NG server has a posting limit), suggesting you take a look

the program MetaCat for an example of how complex behaviour can emerge

the interaction of many simple algorithmic agents (yes, with some

too). You can see it at http://www.cs.pomona.edu/~marshall/metacat/.

isn't about emotions, but it certainly makes a good attempt to model

very human aspects of thinking, particularly aesthetics and

Once you've looked at that in detail, maybe you'll rethink your concept

what we mean by 'algorithmic'.

Nice program, but it doesn't necessarily have much to do with simulating
the human mind. The people who wrote the program would have to
demonstrate that the mind operates on identical principles.

Note too that I'm not saying that AI software run on silicon based
computers couldn't surpass what the mind can do in various limited
areas. I'm not an AI-luddite. What I'm against is AI-arrogance.

Andrew Arensburger wrote:

Earlier in this thread, you wrote:

: I'm interested in the claim from
: those in the computer field that the brain is an **algorithm following
: machine** which can be simulated by a powerful enough computer. I think
: *that* is as big a load of horseshit as anything Penrose has written.

What I'm actually using is the argument from skepticism, which unlike
the argument from incredulity, is not a logical fallacy. In fact,
skepticism is probably the single most popular mode of argument used on
talk.origins.
What's the difference between arguing from incredulity and arguing from
skepticism?

Anyway, my key point against the claim that a completely faithful
simulation of the entire human mind is possible I've already stated:
"we've not even come close to being able to do it." I'll bring this up
again and again with various illustrations. That's how an argument from
skepticism proceeds.

On the other hand, there is the second point of view, the skeptical
view, which goes something like this: "are you f**king kidding me or
what!?" You think it's a big deal that a machine can do what a human
baby does naturally with no effort at all, and can do much better? And
this after at least fifty years of relentless effort by many brilliant
people, during which time the power of computers has grown by at least
one million times? Dude, wake up and smell the coffee. This is a joke,
right?

As an example of our prowess in moving toward the distant day when we
can simulate the whole human mind, I suggest that this is quite dismal
and doesn't auger well at all.

So we've gone from "computers can't play chess" to "the best
human player usually manages not to lose." I think that's respectable.

I don't. What Deep Blue was, was a Gary Kasparov chess playing machine.

Presumably that'll come in a year or two, if Moore's law
holds.

Except that Deep Blue is now retired. However, I have no doubt that if
such chess playing machines continue to be developed, they will
inevitably be able to defeat easily anything human. Brute force will
sooner or later become too much to compete against.

Andrew Arensburger wrote:

In talk.origins dkomo <dkomo871@comcast.net> wrote:

Andrew Arensburger wrote:

If there really is a non-algorithmic component to the brain
that prevents it from being simulated, please point it out, and show
why it can't be simulated. So far, I don't see any problems that can't
be solved by a universal Turing machine with infinite tape, a
microscopic scan of a brain, and an awful lot of programming.

Now wait just a minute. You're making some really far fetched claims
here. I don't see why I need to show why the brain can't be simulated.

Earlier in this thread, you wrote:

: I'm interested in the claim from
: those in the computer field that the brain is an **algorithm following
: machine** which can be simulated by a powerful enough computer. I think
: *that* is as big a load of horseshit as anything Penrose has written.

It's your responsibility to show that it *can*. I can stay pat and
simply say we've haven't even come close to being able to do it.

Actually, you made the positive statement that the brain
couldn't be simulated. You have yet to back up your statement. All
you've done so far is give an implied argument from incredulity.

Strawman alarm! I was anticipating that I'd get accused of using the
argument from incredulity, so I'm not disappointed here.

What I'm actually using is the argument from skepticism, which unlike
the argument from incredulity, is not a logical fallacy. In fact,
skepticism is probably the single most popular mode of argument used on
talk.origins.

What's the difference between arguing from incredulity and arguing from
skepticism? If a creationist or ID'er is involved, then he is using an
argument from incredulity -- against evolution for example. We science
types always use the argument from skepticism against opponents. :>)

Seriously, it might be interesting to hear from some of the logic
experts on t.o. exactly what the differences are, because at least
superficially, there appear to be some similarities between the two
types of argument.

Anyway, my key point against the claim that a completely faithful
simulation of the entire human mind is possible I've already stated:
"we've not even come close to being able to do it." I'll bring this up
again and again with various illustrations. That's how an argument from
skepticism proceeds.

The fact is, after decades of strenuous effort, artificial intelligence
is still pretty brain dead.

Oh? And here I was under the impression that it was a thriving
field with many practical applications, though admittedly it has
fallen short of its original goal, fifty years ago, of producing an
electronic brain.

It can barely achieve the minimum in speech
recognition and visual pattern recognition.

I think that examining a video recording, picking out the
human face in each frame, finding its features (eyes, mouth, eyebrows,
nose, etc.), examining how they move over time and figuring out which
emotion the face is expressing constitutes a bit more than the "bare
minimum" in visual pattern recognition.
(See, e.g., http://www.cfar.umd.edu/~yaser/publications.html )

It's important to distinguish two different ways of viewing such
apparent advances. From the engineeering point of view, I certainly
think that this example of pattern recognition, and others similar to
it, are significant achievements. I know how difficult it is to be able
to do something like this.

On the other hand, there is the second point of view, the skeptical
view, which goes something like this: "are you f**king kidding me or
what!?" You think it's a big deal that a machine can do what a human
baby does naturally with no effort at all, and can do much better? And
this after at least fifty years of relentless effort by many brilliant
people, during which time the power of computers has grown by at least
one million times? Dude, wake up and smell the coffee. This is a joke,
right?

Never mind the human baby! A dog can do as well recognizing patterns --
it certainly recognizes people that are familar to it and responds to
their body language. A cat can do as well. Hell, I bet even a mouse
can do as well. I wouldn't be surprised if most animals in the animal
kingdom can't do as well. A praying mantis recognizes a meal when it
sees one. So do spiders.

As an example of our prowess in moving toward the distant day when we
can simulate the whole human mind, I suggest that this is quite dismal
and doesn't auger well at all.

And the best we've been
able to do in chess playing amounts to little more than than searching
through game trees. A good chess player like Gary Kasparov can still
hold his own against a dedicated chess playing computer that searches
through billions of possible moves every second.

People used to say computers would never be able to do things
that humans do, like play chess. Then they said they'd never be any
good at it. Then they said they'd never be really good at it. Now
you're saying that a grand master can hold his own.
According to
http://www.research.ibm.com/deepblue/watch/html/c.shtml
Gary Kasparov has played six games against Deep Blue. One win for
Kasparov, two wins for Deep Blue, and three ties.
So we've gone from "computers can't play chess" to "the best
human player usually manages not to lose." I think that's respectable.

I don't. What Deep Blue was, was a Gary Kasparov chess playing machine.
It had a dedicated squad of programmers carefully training it for months
on previous Kasparov chess matches. No wonder he lost. It was a
brilliant PR stunt by IBM. They gained prestige and millions of dollars
of free advertising from the exhibition match. Deep Blue has now
morphed into Blue Gene, which I noticed IBM has advertised on television.

What Deep Blue would not have been able to do would be to win the World
Chess Tournament playing against a dozen or so Grand Masters at
Kasparov's level. Deep Blue's programmers would have been overwhelmed
trying to train it to play such a varied assortment of experts. You did
note, didn't you, how quickly IBM decided to retire Deep Blue after the
exhibition match was over and the accolades began to rain down. IBM
knew they would be lucky to repeat their stunt.

I am a lot more impressed that a human being, with a mind that plays
chess using cognitive processes that even today we only dimly
understand, could still compete against such a computerized behemoth, .

And anyway, while I'm not up to date on the state of computer
chess, I understand that there's a lot more to it than just traversing
the move tree. Everyone knows that that's a computational quagmire, so
the trick is to prune unpromising trees as early as possible, which is
what AI is all about.
Oh, and according to
http://www.research.ibm.com/deepblue/meet/html/d.2.shtml
Deep Blue only examines 200 million positions per second, not
billions.

Oh, is that all.

Presumably that'll come in a year or two, if Moore's law
holds.

Except that Deep Blue is now retired. However, I have no doubt that if
such chess playing machines continue to be developed, they will
inevitably be able to defeat easily anything human. Brute force will
sooner or later become too much to compete against.

[End of installment 1]

<snip the rest>


--dkomo@cris.com