Re: "true" AI Hardware Development

Curt Welch wrote:
"JGCASEY" <jgkjcasey@xxxxxxxxxxxx> wrote:
"JGCASEY" <jgkjca...@xxxxxxxxxxxx> wrote:
Curt Welch wrote:

... the brain in fact is not all that amazing. It's not
much more amazing than our computers. The internet is a
machine many times more complex than any human brain.
We have already build signal processing machine far more
complex than anything nature built.

What measure of complexity did you use to compare the brain
and other things that "nature built" with the internet and
current signal processing machines?

CW:
I can't think of a measure that doesn't make the internet
many orders of magnitude more complex than a human brain.

A human brain has 100 billion neurons and maybe 10^15
synapses? How many transistors make up the internet?

A Pentium 4 CPU alone as over 40 million transistors and
ram has one transistors per bit of memory, so a system with
only 512 MB of memory has another 4 billion transistors in
it just for the memory.

One quote I found on the internet says: "By 2001, 268 million
computers will be connected to the Internet". And I believe
that was a reference to end-user machines in 2001 and not
servers. I believe the number of servers connected to the
internet is around 400 million now. But lets just use the
200 million number.

So that's 2e8 computers times maybe 4e9 transistors per
computer giving you 8e17 transistors vs a brain with 1e11
neurons. That's 100,000 transistors per each neuron. And
that doesn't count the fact that transistors switch about
1,000 times faster than neurons.

And that's using very low-ball numbers for the number of
devices connected to the internet today - easily off by a
factor of thousands.

If you look at data flow you find a human brain is receiving
two video signals and two audio signals and a ton of touch
and other data. And in turn, is producing output which is
many orders of magnitude less than what it is receiving.
I've not seen any numbers on what the actual bandwidth of
this information flow is, but I suspect it's less than 100
Mbits. A single PC can beat this easily. The data flow on
the internet is so much larger than this there's no comparison.
Usenet alone has a data flow of around 250 Mbits these days
which means every Usenet server on the planet is pushing more
bandwidth than a human.

Other people estimate the processing power of a brain at
around 100 teraflops but the worlds fastest supercomputer
(Blue Gene) has hit over 300 teraflops now. They will be
over 1000 Teraflops in a year or two. So that single
computer (which is probably connected to the internet)
rates as faster and more complex than a single human brain.

I can't think of a metric to measure the complexity of a
signal processing device that doesn't put the internet as
a whole many orders of magnitude ahead of the human brain.

The amount of complex behavior produced by the entire
internet in a few seconds is far greater than a human
produces in a life time. No matter how you choose to
measure complexity of a single processing device, the
internet will always win by many orders of magnitude
compared to a single human brain. Humans are just slow
stupid machines compared to our computers. Compared
to the internet as a whole, a human brain doesn't even
come close.

The brain remains special not because of it's raw power,
but simply because of what it can do. Before long, we
will figure out how to build machines that do the same
thing, and when that happens, the brain will look just
as small and weak as a human arm looks small and weak
compared to a robotic manufacturing arm.

------------------------

JC:

You seem to equate "complexity" with quantity, strength
and speed?

Yes.

You talk about the "processing power" in terms of teraflops
but that doesn't mean the processing is complex only that
it can be done really fast.

Imagine two cellular worlds with the same amount of cells.
Give one world Conway's rules and the other another set
of rules. Same quantity of cells but I suspect the result
will vary in complexity.

You might say that the harder it is to predict or describe
"what the system is doing" the more complex it is?

The internet may be large but that doesn't mean it is doing
anything complex.

In terms of possible states I think the brain probably isn't
all that complex and that may be why it works. Too much to
deal with and nothing is decided. Break the problems into
simpler less complex forms and they become solvable.

The "complex" interaction of water molecules at the bottom
of a waterfall isn't going to do anything "intelligent" despite
the quantity of water molecules involved.

Yes, but we weren't talking about intelligence where we?

I assumed you were talking about a relationship between
intelligence and complexity.

We were talking about complexity. And the complexity of what
happens at the bottom of a waterfall is far more complex than
what humans do.

But the resulting macroscopic behavior is not complex. The complex
interactions of the molecules of a gas can be reduced to the "gas
laws" based on measuring temperature, volume and pressure.


The reason I talked about complexity is that you can't duplicate
intelligent behavior until you first duplicate it's level of
complexity. A computer with 1K of memory will never produce the
complexity of behavior of Microsoft Windows. It can't be done.
You need a machine with the power to produce more complexity.
But just because you have a machine with enough power, doesn't
make it Microsoft windows any more than expecting the internet
to act intelligent.

We don't really know how much complexity is required to produce
what level of intelligence.

Essentially the brain is a regulator that helps to maintain
the essential variables that define the dynamic system we call
an animal. The brain blocks any environmental disturbances
that may put the animal into a fatal unstable state.

The gene patterns most likely to survive would have placed a
defense between them and any environmental factors that may
destroy them. This begins at the cellular level but at the
multicellular level the genes coded for passive defenses like
a shell or active defenses like a brain.

How well the animal's actions protect the animal depends on
having a action to counteract every environmental action that
may bring the animal undone. So the complexity of the
behaviors would have to match the complexity of the forces
acting against the animal.

Getting food, avoiding being food and reproducing are the
name of the game and the foundation of all our behaviors.


My comments about complexity are just meant to stress the fact
that we have already build machines far more advanced and
complicated than the human brain. We just haven't built the
correct complex machine yet to duplicate human intelligent
behavior.

But they are not "more advanced" in doing the things the
brain does well and we haven't decided on how to measure how
"complicated" something has to be to duplicate the human
brain. I suggested it depends on the complexity of the
environment and I would also suggest it is more a matter
of quality than quantity when it comes to intelligence.

Interacting water molecules and internets may be complex but
I suspect they are dumb as ....

--
JC

.

Relevant Pages

  • Re: "true" AI Hardware Development
    ... We have already build signal processing machine far more ... and other things that "nature built" with the internet and ... many orders of magnitude more complex than a human brain. ... computers will be connected to the Internet". ...
    (comp.ai.philosophy)
  • Re: "true" AI Hardware Development
    ... and other things that "nature built" with the internet and ... many orders of magnitude more complex than a human brain. ... computers will be connected to the Internet". ... that was a reference to end-user machines in 2001 and not ...
    (comp.ai.philosophy)
  • Re: "true" AI Hardware Development
    ... machine many times more complex than any human brain. ... and other things that "nature built" with the internet and ... One quote I found on the internet says: "By 2001, 268 million computers ... end-user machines in 2001 and not servers. ...
    (comp.ai.philosophy)
  • Re: Exploiting limitations of Turing machines in Turing tests?
    ... the neurons of an human brain. ... but I'm not too sure if such a simulation is possible. ... computers to keep our computers from choking on the data the brain deals ... The human brain can't touch it. ...
    (comp.ai.philosophy)
  • Re: AAT yahoo group more than 500 members
    ... imitation, and those that replicate via computers, copiers, etc. ... concepts are manufactured in the human brain. ... because tunes are not replicators while they are in the ...
    (talk.origins)