On Sun, Aug 25, 2013 at 5:11 PM, John Clark <johnkcl...@gmail.com> wrote:
> On Sat, Aug 24, 2013  Telmo Menezes <te...@telmomenezes.com> wrote:
>
>> > We are now approaching a point where we can have supercomputers with the
>> > same estimated computational power of a human brain, but we are very far
>> > from replicating its capabilities.
>
>
> Very far?

It think so, in terms of generality of ability to learn and
adaptability to new situations.

>
>> >  Chess is a very narrow case
>
>
> But being the best Jeopardy player in the world is far less narrow,

Yes, and Watson is very cool, but still not a human-like mind.

> and in
> the current issue of New Scientist is a article about an AI program running
> on a computer MUCH smaller than Tianhe-2 that scored what a typical 4 year
> old child would on a verbal IQ test.

Doesn't that just involve screaming inane things all the time? :)
But seriously, I'll have a look and comment later.

>>
>> > Also, Moore's law is bound to hit a physical limit. It cannot be that
>> > far now.
>
>
> Fans of innate human superiority have been singing that same tired old song
> for 40 years.

Ok. I'm not a fan of human superiority. My motivation is a desire for
the type of AI I dream about. I'm confident it can be achieved, but
what we have now doesn't satisfy me.

> Currently the smallest features on the very best computer
> chips on the market are about 22 nanometers and the chips are primarily 2D.
> There is no physical reason you couldn't make 3D logic elements with
> features as big as medium sized molecules or about 1 nanometer, so per
> volume you could pack 22^3 = 10,648 more logic elements than what is
> currently the most complex and tightly packed object (the surface of a
> microprocessor) humans have ever made. And because the elements are 22 times
> closer together they could transfer signals 22 times faster, so per volume
> you could solve problems 22^4 = 234,256 times faster.

My understanding is that heat dissipation becomes a problem below 22
nanometers, but I'm not an expert. I assume you're a physicist so
sure, you tell me. I did take a course in computer architecture more
than a decade ago, and I remember there were two issues with things
like smaller scales and more layers: heat dissipation and clock
synchronisation. But again, I take no pleasure in the end of Moore's
law, quite the contrary and I hope you're right.

> And after that is
> quantum computing.

Sure. We hope.

> There's more, a lot more. The fastest signals in the human brain move at a
> couple of hundred meters a second, many are far slower, light moves at 300
> million meters per second. So if you insist that the 2 most distant parts of
> a brain communicate as fast as they do in a human brain (and it is not
> immediately obvious why you should insist on such a thing) then parts in a
> AI's brain could be at least one million times as distant. The volume of
> such a brain would be a million trillion times larger than a human brain.
> Even if 99.9% of that space were used just to deliver power and get rid of
> waste heat you'd still have a THOUSAND TRILLION  times as much volume for
> logic components as humans have room for inside their heads, and per volume
> the components would be solving problems enormously faster than anything we
> have today.

Yes, this will be great, but we don't know how to program massively
parallel asynchronous machines like this. Not saying it can't be done,
just that it's a problem we still have to solve. I'm actually quite
interested in this problem and have a couple of ideas that involve
self-modifying computer code and embedded evolution. But nothing to
show for it so far.

Telmo.

>   John k Clark
>
>
>
>
>
>
> --
> You received this message because you are subscribed to the Google Groups
> "Everything List" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to everything-list+unsubscr...@googlegroups.com.
> To post to this group, send email to everything-list@googlegroups.com.
> Visit this group at http://groups.google.com/group/everything-list.
> For more options, visit https://groups.google.com/groups/opt_out.

-- 
You received this message because you are subscribed to the Google Groups 
"Everything List" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to everything-list+unsubscr...@googlegroups.com.
To post to this group, send email to everything-list@googlegroups.com.
Visit this group at http://groups.google.com/group/everything-list.
For more options, visit https://groups.google.com/groups/opt_out.

Reply via email to