On Tue, Aug 16, 2011 at 9:32 AM, benjayk <benjamin.jaku...@googlemail.com>wrote:

>
>
> Jason Resch-2 wrote:
> >
> > On Tue, Aug 16, 2011 at 7:03 AM, benjayk
> > <benjamin.jaku...@googlemail.com>wrote:
> >
> >>
> >>
> >> Craig Weinberg wrote:
> >> >
> >> > On Aug 15, 10:43 pm, Jason Resch <jasonre...@gmail.com> wrote:
> >> >> I am more worried for the biologically handicapped in the future.
> >> >>  Computers
> >> >> will get faster, brains won't.  By 2029, it is predicted $1,000 worth
> >> of
> >> >> computer will buy a human brain's worth of computational power.  15
> >> years
> >> >> later, you can get 1,000 X the human brain's power for $1,000.
> >> Imagine:
> >> >> the
> >> >> simulated get to experience 1 century for each month the humans with
> >> >> biological brains experience.  Who will really be alive then?
> >> >
> >> > Speed and power is for engines, not brains. Good ideas don't come from
> >> > engines.
> >> >
> >> > Craig
> >> >
> >> I agree. It is a very narrow to think computational power is the key to
> >> rich
> >> experience and high intelligence. The real magic is what is done with
> the
> >> hardware. And honestly I see no reason to believe that we somehow we
> >> magically develop amazingly intelligent software.
> >
> >
> > Neural imaging/scanning rates are also doubling every year.  The hope is
> > that we can reverse engineer the brain, by scanning it and making a map
> > all
> > the connections between the neurons.  Then if the appropriate hardware
> can
> > run a few brains at 1,000 or 1,000,000  times faster than the biological
> > brain, we can put our best scientists or AI researchers inside and they
> > can
> > figure it out in a few of our months.
> >
> > http://www.kurzweilai.net/the-law-of-accelerating-returns
> There are *so* many problems with that. We are naive, a bit like 7 year old
> wanting to build a time machine. We know little about the brain. Who says
> there is no quantum effects going on? There doesn't even have to be
> substantial entaglement. Chaos theory tells us that even minuscle quantum
> effects could have major impacts on the thing. ESP and telepathy suggest
> that we are to some extent entangled. There are *major* problems
> reprodocing
> this with computers.
>
> Neural imaging and scanning cannot pick up the major information in the
> brain. Not by a long stretch.


Automated serial sectioning of brains is already fairly advanced, and is
doubling in performance and accuracy each year.
http://www.mcb.harvard.edu/lichtman/ATLUM/ATLUM_web.htm


> It is like having a picture of a RAM and
> thinking this is enough to recover the information on it.
>
> What use are fast brains?


A million years of human technological progress in the time frame of one
year seems highly useful.


> Our brains alone are of little use. We also need a
> rich environment and a body.
>

I'm not sure bodies are necessary, but in the context of a simulation you
could have any body you wanted, or no body at all.  (Like in second life)


>
> You presuppose that AI researchers have the potential ability to build
> superintelligent AI. Why should we suspect this more than we suspect that
> gorillas can build humans? I'd like to hear arguments that make it
> plausible
> that it is possible to engineer somthing more generally intelligent than
> yourself.
>

I there was someone just like me, but thought at twice the speed, I am sure
he would score more highly on some general intelligence tests.  If we can
find a gene or genes that make the difference between Newton and the average
person, and then switch them on in the average person through gene therapy,
would that count as engineering something more intelligent than yourself?
What about taking Nootropics ( http://en.wikipedia.org/wiki/Nootropic )?
There are many plausible scenarios for making ourselves more intelligent, or
more creative than our current state.


>
>
> Jason Resch-2 wrote:
> >
> >> Software development is
> >> slow, no comparison to the exponential progress of hardware.
> >>
> >
> > As I mentioned to Craig who complained his computer takes longer to start
> > up
> > now than ever, the complexity of software is in many cases outpacing even
> > the exponential growth in the power of computer hardware.
> That may quite well be. But even if we have a software that can render a
> 99^99 dimensional mandelbrot this will not be of much use. The point is
> that
> the usefulness of software is not progressing exponentially.
>
>
> Jason Resch-2 wrote:
> >
> >> I believe that it is inherently impossible to design intelligence. It
> can
> >> just self-organize itself through becoming aware of itself.
> >
> >
> > A few genes separate us from chimps, and all of our intelligence.
> I don't think our intelligence is reducible to genes. Memes seem even more
> important. And just because we can't really research it scientifically at
> moment, does not mean there are no subtler things that determine our
> general
> intelligence than genes and culture. Many subjective experiences hint at
> something like a more subtle layer, call it "soul" if you will.
> All of what we understand about biology may just be the tiny top of a
> pyramid that is buried in the sand.
>
>
Well we only have one sample of biology from one planet in one type of
chemistry.  Throughout the everything, what we know is minuscule compared to
what can be known about biology.  That said, there is a finite amount there
is to learn about the human brain and biology.  There are information
theoretic limits established by the number of base pairs which set upper
bounds on how much there is to be learned about human biology.  With this,
we confidently say the brain's design is not infinitely complex.



>
> Jason Resch-2 wrote:
> >
> >   If we can
> > determine which, and see what these genes do then perhaps we can
> > extrapolate
> > and find out how our DNA is able to make some brains better than others.
> But this is not how intelligent works. You don't just extrapolate a bit and
> have more intelligence. If this were the case, we would already have
> superintelligence.


Vastly super-human intelligence requires vastly more powerful processing
capabilities.  Our computers are still very far from that, and are more on
the level of insect brains.


> Development / evolution of intelligence, learning and
> consciousness are highly non-trivial, and non-linear.
>
>
> Jason Resch-2 wrote:
> >
> >> I am not even
> >> sure anymore whether this will have to do very much to do with
> >> technology.
> >> Technology might have an fundamental restriction to being a tool of
> >> intelligence, not the means to increase intelligence at the core (just
> >> relative, superficial intelligence like intellectual knowledge).
> >>
> >
> > I think the existence of Google and Wikipedia makes me more intelligent.
> > If
> > I could embed a calculator chip into my brain my mental math skills would
> > improve markedly.
> This is exactly the kind of intelligence I am NOT talking about. It's
> useful, sure. But it doesn't lead to unimaginable creative, self-improving
> intelligence. We may become super-knowledgable in the next decades, sure.
> But this doesn't mean we acquire the wisdom that is necessary for deep
> progress, leading to higher states of consciousness and happiness.
> Whether you know extremely much or not isn't really so important that it
> would divide enhanced humans from normal humans more than it divides
> scientifically literate people from badly educated people.
>
>
>
> Jason Resch-2 wrote:
> >
> >>
> >> Also, we have no reliable way of measuring the computational power of
> the
> >> brain, not to speak of the possibly existing subtle energies that go
> >> beyond
> >> the brain, that may be essential to our functioning. The way that
> >> computational power of the brain is estimated now relies on a quite
> >> reductionstic view of what the brain is and what it does.
> >>
> >
> > As I've mentioned before on this list, neuroscientists have succeeded in
> > creating biologically realistic neurons.  The CPU requirements of these
> > neurons is well understood:
> >
> > http://www.youtube.com/watch?v=LS3wMC2BpxU&t=7m30s
> Biologically realistic neurons is relative. We certainly don't take quantum
> effects into account, and evidence seems to suggest this is important.


On the contrary, the simulated neurons behaved in the same ways as
biological neurons, and this is without including quantum effects.  It would
therefore seem that quantum effects play a negligible role in a brain's
function.


> At
> least I see no other way to explain ESP.
>

I don't see how quantum mechanics could even explain ESP.  Entangled
particles cannot be used to transmit information.


>
> Even if we suppose that they are biologically realistic, neurons alone
> don't
> make up human a functioning brain. A neuron is like a transistor of a
> computer, and a transistor is not enough for a functioning computer! There
> are other type of cells that may be important in information processing.
> Also there are different kinds of neurons, and the way they are put
> together
> in different units is also important. Even if we were able to reproduce all
> of this, we would still need the software running on the brain.
> How would we do this?
>

The software of the brain is represented by the manner in which the neurons
are connected to each other, and how individual neurons respond to each
other.  This design can be copied straight from the data provided by serial
sectioning scanning.


> At the very least, it seems we would have to biologically realistically
> simulate the development of the brain from before birth until late
> childhood. And of course this needs a very good interface to the outside
> also.
>

I think it will be possible to take a frozen brain from an adult human, scan
its synapses and structure, load a model of this brain structure into a
computer, and then emulate how the brain would respond to inputs.


>
> So, simulating neurons seems to be the easiest task in simulating a brain.
>
>
>
Once we have that, the remaining two problems are: How to scale the
simulation of a single neuron to a hundred billion neurons and how the
neurons should initially be connected to each other.

Jason

-- 
You received this message because you are subscribed to the Google Groups 
"Everything List" group.
To post to this group, send email to everything-list@googlegroups.com.
To unsubscribe from this group, send email to 
everything-list+unsubscr...@googlegroups.com.
For more options, visit this group at 
http://groups.google.com/group/everything-list?hl=en.

Reply via email to