Re: Recycling/AI and super human computers

[T.L.G.vanderLinden@student.utwente.nl (Timothy van der Linden)]

>What I was going for was that the TV's them selves have changed dramatically
>without HDTV.  These changes in image quality and relyability were enough
>that everyone went out to get a new set.

I'm not so sure about that, a lot of people only buy a television because
their old one breaks down and the repair cost are almost as high as a new
TV-set.
Only things like remote control or the number of programmable channels are
features that may tempt people to buy a new set while their old one is still
working.

>I wounder if the internet will wind up becoming part international cable
>television?

As far as I know all dataflow will be incorporated into a "single" network
(eg. ISDN). The communication-protocol is the important thing her, the
physical form of the network just has to assure that enough bandwidth is
available.

So it's more the other way around, cable-tv is becoming part of internet.
However it may look the otherway around since cable-companies will become
providers of internet. What I mean is that internet already is a
multi-purpose network, it can be used for text, images, audio, movies. The
thing that people are working at today is to get a protocal that has two
parts: a steady-stream fixed bandwith and a burst-stream non-fixed bandwith.
The first will can be used for live audio and video while the latter can be
used for email, ftp or normal internet use.

>Less than that.  Computer systems have maintained a 100 fold improvement in
>performance per decade for over half a century.  Hard to tell how A.I. would
>fit into that.  We might make a breakthrough (like in my book) and have
>everything fall into place in months, or we might stumble along for decades.

I go for the few months...

>>> Maybe we should make them just smart enough to do 
>>> the dirty jobs. And use only a few with IQ 1000 to think. 
>
>;)  You can't hide forever.  Best to come out a deal with things as theyt
>happen, not try to lock yourself into a safe past.

It's not lock yourself, it's protecting yourself. It's like saying, why not
do any kind of genetic experiment, you can't stop it anyway.
Of course smart computers are not necessary dangerous, maybe they LIKE to
help us out. If they are convinced that we are not just annoying bugs, then
the problem may not be that worse.

>>> Yes, I found some info about another CRAY having max 1.2 TFLOPS.
>
>Ok, so we have the electronic brains.  Now for the minds!

As soon as the structure of the brain and its input is ready the mind will
follow automatically.
The input is something that a lot of people forget. I think for any brain to
start working, the input should be sufficiently high. And if we want a
computer that is least likely to kill us and most likely to understand us,
we should make its senses similar to ours.

I think there is another possibility to control the AI, we let it "live" in
a virtual reality, which is created by us. So every action the AI undertakes
will not be a real one (so no harm to us), but since we can control it's
input it may never know that it is not real.

In fact, we don't know that we are real. Of course this depends on what one
calls real. For most of us everything that we can measure is real, but there
is that stuff where dreams are made of, why not call that real too?
In short we are not able to figure out if there is some smart superhuman
that is controlling our senses. How can you be sure that you are not in a
never ending dream, creating everything and everyone all by yourself?

>> Custom circuis are is less efficient to do highly parallel computing than
>> neural-circuits. ---
>
>I was refering to neural net circuts.  We alread know how to make them to the
>same consentration as standard circuts, but so far they arn't as usfull.

Not as useful as we would like...

Timothy