Re: How Robots Will Steal Your Job

[Roedy Green]

Well then the computer will never be intelligent, because it
can't home in on the solution by itself, it has to have
another "species" in the loop, either so it can measure the
other species reaction (which is a result of judgement on
the other species part), or it needs the other species to
tell it what a desirable solution is. By itself, it's just
a dumb bunch of parts.

Intelligence will always be defined so that only humans have it. It
is bit like having a soul. As computers evolve, we will have to go
through greater and greater contortions to defend our superior
position.

It is John Henry by the thousand cuts. Computers are already better
than humans at designing electric transmission lines, diagnosing rare
diseases, face recognition with disguises, memory, assembly line work.

A book that might shake you up a bit about the ultimate potential for
computer intelligence is Ray Kurzweil's Spiritual Machines. He is the
guy who invented modern OCR, holds sythesiser patents, and a futurist
with a track record that beats the rest.
http://www.amazon.com/exec/obidos/ASIN/0140282025/canadianmindprod

see http://mindprod.com/heroes.html#KURZWEIL

We go to great lengths also to convince ourselves we are more
intelligent that dolphins and whales. Since we control the definition,
it makes it a lot easier to do. See http://mindprod.com/intel.html

This leads to a values crisis. North Americans gain their value by
their work. They may not be able to continue doing that much longer.
See http://mindprod.com/work.html

 

[Roedy Green]

On the other hand, our brain is three or four times the size of
a chimpanzee brain, and this size increase happened in only
about 4 million years, which is astounding. There must be
something we can do that the chimps can't. We build cities and
spaceships.

Raw size is just a cheap trick. We humans like to pretend we have
some complex circuitry unlike anything else in the animal kingdom.

Changing size is something evolution handles with ease. Devising new
structures is something else.

In computers, there is quite a difference in what a 4K computer and a
4 gig computer can do. However, It would not be obvious if you had
only seen a 4K computer.

Up to now, human had a terrible speed disadvantage and a massive size
advantage over computers. That size disadvantage is rapidly being
whittled away. Computers also have the advantage of deliberate
design. The human is a kludge that sort of worked. If you don't
believe me, ask a woman in labour.

 

[Richard Heathfield]

Yes, AND it is a problem solving process.

Whose problem?

If it mutated too slowly
there would not be enough natural variation to select from.

The words "not...enough" imply that someone or something cares, and is in
control. Who?

If it
mutated too fast too many individuals would die young.

"Too fast" for whom?

Ask a Thai how to develop a fish with longer fins and he will say,
"kill off the ones with short ones".

In other words, you pre-suppose an intelligent agency.

This is the same strategy nature
uses to inadvertently create ever creatures ever more suited for the
environment.

The word "inadvertently" is key here. Either there is an intelligence
guiding evolution, or there is not. If there is not, then your words make
no sense. If there is, then the process of evolution is not random.

Check out http://mindprod.com/jgloss/montecarlo.html

You appear to have missed the point I am making, or you would not have
referred me to that page.

 

[Roedy Green]

These things, although complex and involved, are deterministic and
have specific algorithms for solving. Things requiring creativity
or insight are a whole different class of problem!

there are programs that paint, not just abstracts.

 

[Roedy Green]

I dunno about that. How can we build a machine that can supercede the
human brain ?

Same way it is possible to build one that can "run" faster, or "lift"
more weight, or "see" farther, or "hear" quieter sounds.

 

[Roedy Green]

If the human brain was simple enough to understand, we would be too
simple too understand it (this may already be the case). All we know
about the brain is its tissue composition.

It was Einstein who was astounded that the universe was as
comprehensible as it was. It is very nice of it to be so simple that
even monkeys like us can figure out the basic principles.

However, having created an AI program of sorts, and seen it do things
that surprised ME, I have much higher hope for AI generally.

What passes for intelligence can be faked as, and likely is, a lot of
bullwork, dull computation.

Note that it takes time to come up with a new solution, perhaps on a
walk in the park. You are often not even aware of the process by
which the answer appears.

I'm surprised that with the success of neural nets at face recognition
many still believe that computers are only big calculators.

 

[Roedy Green]

docs for the human brain, we are, by definition, not smart enough
to figure it out.

We are almost to the point where you could microslice and digitise a
brain so that you had the digital anatomy and connections of all the
cells. That is not that much further off than sequencing DNA seemed a
decade ago.

 

[Roedy Green]

its not just about the volume of information. its about how it is used.

presumably the human explores deeper but with less breadth. It has
better patterns for pruning unpromising branches.

 

[Roedy Green]

Brute force searching is not intelligence.

If you thought a problem was beyond the range of brute force
searching, you would say that anyone or anything that solved it,
namely you, had to be intelligent.

You may be unaware that is how you are internally solving it.

If someone later solves it and lets you peak in the black box and you
see it is a brute force search of likely regions of solution space
using special purpose hardware to help, then you decide this no longer
counts as intelligence. You move the goal posts.

Unless it is mysterious and something only I have, it can't be called
intelligence. This is a sort of religious worship of homo sapiens.

 

[Airy R Bean]

The search for revelation in the matter of Artificial Intelligence
will not be resolved by silly and childish arguments about on which
side of a piece of toast should be buttered.

There are many in this field who seek to be perceived as experts.
Perhaps it is their complete lack of expertise that they hope to
dissemble by these spurious arguments?

Airy R Bean wrote:
Misquoting corrected.....
[Top-posting fixed]
Airy R Bean wrote:

(etc!)
DFTT Richard. I know it's annoying, but just ignore him.

Your wish is my command, Corey. Life's too short.

 

[Airy R Bean]

The domestic cat clearly has intelligence from the way
in which it resolves barriers that are placed in the path
of its nighly prowl.

However, that is no evidence of self-awareness. We will
not find the solution to our quests by confusing the issues.

 

[Richard Heathfield]

But, presumably it is not. Olaf Stapeldon might agree with you, but
the general presumption in our culture is that things unlike us are
not conscious.

It may, or may not, be the general presumption. I see no evidence either
way. Nevertheless, my point was in line with the general presumption, in
this case, and was designed to show that words like "desirable solution"
imply that something is consciously doing the desiring. I did not intend to
imply that astronomical bodies have consciousness (although I have no
evidence to the contrary); rather, I was highlighting the careless wording
of the argument to which I was responding.

That is an interesting discussion in itself. Given that we have no
way yet to measure consciousness, it is all just guesswork.

Right.

 

[Programmer Dude]

If course. However, if you DO store them that way then much faster
and more sophisticated tools COULD potentially be available.

Heh. Kind of a chicken/egg problem.

Until such tools do become available, so long as you had
export/import, you could use the traditional ones too.

Right (although that does raise the versioning issue). All in all,
I still think I *very* much prefer one source (and that source is
text).

There's also the consideration of sharing source, and that text is
the least common denominator format. I have some VB3 files that
I can probably never access again. (VB used a "special" binary
format for source until VB4 when they went to text.)

 

[Programmer Dude]

Perhaps what we mean by human intelligence in the ability to
generalise.

That is certainly a huge part of it. My dog has a set of distinct
round toys--each is a separate item with its own "personality". To
me, it's just a bunch of tennis balls (all came from the same
container so all have the same markings and age).

It can be really funny. When we're playing, I'll grab one of the
other balls when she can't see and then switch them when I get
possession of the current play toy. I'll put the "wrong" one in
play ... and she's NEVER fooled. She'll totally ignore it.

 

[Programmer Dude]

Therefore, if "gain mass" is a desirable solution, an astronomical
body can show intelligence merely by switching on gravity.

Eh? One gains mass by switching on gravity?

Since all astronomical bodies (modulo weird exceptions, of course)
/have/ switched on their gravity, we can deduce that they are all
attempting to gain mass intelligently.

Bizarre science aside, to come to that conclusion, you'd have to
first establish the truth of your leading IF.

Any squirrel can generalise. ("Hmmm, looks like a nut, cracks
like a nut, smells like a nut. Must be a nut.")

As I understand it, it's quite the opposite of generalizing. I'd
bet dollars to donuts the squirrel sees each nut as distinct, but
simply recognizes food when it discovers it.

Generalizing would lead to counting, and we're pretty positive
that squirrels don't count.

You don't suppose "thinking" might be a more significant part of
intelligence than "generalising"? Hmmm?

Oh, I DO suppose! (-:

 

[Programmer Dude]

How do you know?

Observation (plus it's the current "party line" on the matter).

Why do you discount obvious learning (e.g. Pavlovian responses)...

Conditioned response has nothing to do with a time sense. It's
a re-wiring of the brain (as is all learning). It's nothing more
than an adjustment to the neural net so that a given stimulus has
a higher probability of generating the response. Nothing to do
with a time sense.

...and caching (squirrel knows winter will be hard, so lays in a
food supply) behaviour?

As far as we know, it's purely hard-wired instinctive behavior.

Again, how do you know?

Observation. No animal I've ever known (and I've known a few) has
ever demonstrated that sort of curiousity. Nor have I ever heard
of such. I would assume that if animals were capable of that level
of abstract thought, we'd see *some* sign of it.

I've often wondered what goes through an animal's mind, but I
don't actually /know/.

So have I, and likewise!

What magic insight do you have?

Nothing magical, just my observations. (I like most animals better
than I like most humans, so I've been around them a bit, and I've
payed a lot of attention to them.)

 

[Programmer Dude]

there are programs that paint, not just abstracts.

From what (little) I've seen, they produce very derivative--and
not very interesting--works.

 

[Programmer Dude]

How can we build a machine that can supercede the human brain ?

Same way we build machines that supercede human strength and
motive power. Machines already supercede the human brain in
some regards (perfect recall, sheer speed, etc.).

All we know about the brain is its tissue composition.

We know a *little* more than that, although we have a long way
to go. (Was just reading about new thoughts about the role played
by the cerebellum, for example.)

...but unless someone hands us the blueprints and design
docs for the human brain, we are, by definition, not smart
enough to figure it out.

I think we'll eventually figure out those blueprints ourselves.
I don't see any sense of "by definition" in that. Some of it is
just a bioengineering problem requiring better tools than we now
have, but I do think we'll get there (we ARE getting there!).

Once we do, we should be able to solve the great question about
consciousness--is it merely an emergent property of complexity,
or does it require something ineffable we can't build. It may
turn out that we can build a brain, but only God can build a mind.

getting a computer to *think* is kinda like getting a submarine
to swim (or something like that .

"The question of whether a computer can think is no more
interesting than the question of whether a submarine can
swim." (Edsger W. Dijkstra)

computers can only make sense of something if *we* can make sense
of it, because *we* tell the computer what to do.

For now. That may change. We already use computer programs to
design other computers. Imagine that after a few levels of
indirection!

 

[Programmer Dude]

We humans like to pretend we have some complex circuitry unlike
anything else in the animal kingdom.

We clearly have something. No *other* animal has taken over the
planet (nor come close to messing up the entire biosphere).

Computers also have the advantage of deliberate design. The
human is a kludge that sort of worked.

Self-evidently worked pretty well. There is a big advantage to
being non-specialized.

 

[Programmer Dude]

Just shift your emphasis to "noticing" or "finding" patterns, and
you have the distinguishing characteristic of intelligence.

I would prefer to call it "*A* distinguishing characteristic..."

What humans are so good at is noticing that some aspect of their
experience is recurring, which is to say that there is a pattern
to be found, that there is something predictable about their
experience. [...] Pretty much all animals do this to some extent.

Yes. My dog is *amazing* to me sometimes in her ability to notice
patterns. E.g. I have a good friend who comes over every Wednesday
night, and we go to the movies. I am beginning to suspect my dog
has discovered a pattern in what I do Wednesdays between getting
home and his arrival, because she's started to be very alert about
the front door.... on Wednesdays. Whatever pattern she's picking
up on, it must be fairly subtle, because there just isn't that
much different in my patterns AND there is considerable variation.

OTOH, humans are *so* amazing at pattern detection, we frequently
find patterns where none exist. And, as you say, we are far above
the critters in detecting abstract patterns.

Once we get that general pattern-finding mechanism implemented in a
machine and functioning at human levels, it will no longer appear
as a "cheap trick". Our machine will be able to learn anything a
human can learn.

I think there's more to it than that. We aren't just good at
detecting patterns, we're good at *synthesizing* explanations.
(Consider epicycles in early astronomy!)