Re: brain thoughts

casey <jgkjcasey@xxxxxxxxxxxx> wrote:
On Nov 29, 2:06=A0am, Wolf K <weki...@xxxxxxxxxxxx> wrote:
casey wrote:
On Nov 28, 12:25 pm, "rs...@xxxxxxxxxxxx" <rs...@xxxxxxxxxxxx> wrote:
...
We live today in the century of molecular biology.

It is unclear to me what your interest here is?


Artificial intelligence.

Which is not congruent with programs running on digital
machines. Not even programs simulating neurons (though
such programs may be useful for learning about intelligent
machines.)

A simulated cyclone cannot blow down a house but a robot
can fetch a can of beer. I don't see AI as a simulation.

A lot has to do with your notion of intelligence.

I don't see "intelligence" in the singular. There is no
one solution. It is a category word for certain kinds
of behavior.

A meat computer is much more efficient.

Not relevant. The efficiency is either efficient enough
or it isn't for any given task.

JC

If efficiency was a problem, then yes, we might not be able to equal _full_
human behavior with computer parts because it might take more power than we
have available on the earth to drive it. But if that was the only thing
that was stopping us, we would still be able to build giant power hungry
machines that demonstrated simplistic animal-like behaviors - which no one
has yet done at any power level. It just seems totally invalid to argue
that efficiency has anything to do with the problem at this point. It's
just not a lack of power that's keeping our robots from acting like humans.

From most ways of measuring it, computers and electronics are far more
efficient than biological systems anyhow. Transistors can switch with far
less energy than what it takes to make a neuron fire.

The problem is that it's a type of machine we don't yet understand how to
build. As I just wrote in the other post, I think it's a machine that can
better be seen as a chaotic behavior generator that works in parallel but
which gets conditioned with a global reward signal that biases the path
though state space the machine likes to travel - where the "path though
state space" is the combined internal signals and outputs it produces in
response to the stimulus signals. It's got to be some type of constrained
chaotic system where the constraints get adjusted by conditioning. My
networks have all been attempts to create such systems.

--
Curt Welch http://CurtWelch.Com/
curt@xxxxxxxx http://NewsReader.Com/
.

Relevant Pages

  • Re: Can you learn computer science from a school?
    ... But this of course is just arguing my original point, that the real-world constraints of real-world machines necessarily influence algorithm design, and so Computer Science is concerned with these real-world constraints as well. ... First, "efficiently" computable isn't quite the right word, because efficiency typically goes to the resources required, compared to some ideal, rather than "can I run this computation in less than the expected life of my computing system?". ...
    (comp.lang.lisp)
  • Re: Budget Linux cluster suggestions?
    ... the thermal IR (30 THz) rather than the near-IR. ... Look at the top 5 machines on the TOP500 supercomputer list and compute the efficiency. ... This gives an indication of how much of the potential compute power can actually used in applications. ...
    (sci.electronics.design)
  • Re: Pinball vehicle
    ... But never for fuel efficiency. ... For that in a pinball ... diesel cars and small trucks due to the tax structure on ... again I don't haul machines around long hauls but on occasion. ...
    (rec.games.pinball)
  • Re: Toroid for mains conversion?
    ... The percentage efficiency of similar rated E-cores and toroiodals is ... would be in favor of the toroid. ... and the toroid-winding machines were _always_ ... used ferrite pot core transformers for audio... ...
    (sci.electronics.design)
  • Re: Toroid for mains conversion?
    ... The percentage efficiency of similar rated E-cores and toroiodals is ... would be in favor of the toroid. ... and the toroid-winding machines were _always_ ... Isn't the object of a torroidal transformer greater isolation and less ...
    (sci.electronics.design)