How Long Have We Had Petaflop Ability?

[INRM]

How long have we had computers that could run into the petaflop speed range?

[paximperium]

Here's what I did.
I copied and pasted "How long have we had computers that could run into the petaflop speed range?" onto Google and here is the very first result:
http://blog.wired.com/wiredscience/2009/02/supercomputer.html

Do you have a problem using google or something?

[Marquis de Carabas]

I'd say the ability was there since 1995 or so, but the first real petaflop was probably in 2003 (http://www.imdb.com/name/nm0933959/).

[erlando]

Here you go (http://tinyurl.com/2j6bh)

[zooterkin]

How long have we had computers that could run into the petaflop speed range?

The Forum server was upgraded only a couple of weeks ago, but I didn't realise it needed to be that fast..

[Dr H]

Here's what I did.
I copied and pasted "How long have we had computers that could run into the petaflop speed range?" onto Google and here is the very first result:
http://blog.wired.com/wiredscience/2009/02/supercomputer.html

Do you have a problem using google or something?

"With 20 petaflops of computing power, meteorologists could predict local weather down to the 100-meter range. For an event like a tornado, that could mean being able to predict the path that the twister takes through a town, allowing for targeted evacuations that save lives. "

"When it's installed at Lawrence Livermore National Laboratory in 2012, it could make new kinds of calculations possible, but initially, that power will be primarily used to simulate nuclear explosions, as many of its supercomputer forebears have done."

Sigh.

[paximperium]

"With 20 petaflops of computing power, meteorologists could predict local weather down to the 100-meter range. For an event like a tornado, that could mean being able to predict the path that the twister takes through a town, allowing for targeted evacuations that save lives. "

"When it's installed at Lawrence Livermore National Laboratory in 2012, it could make new kinds of calculations possible, but initially, that power will be primarily used to simulate nuclear explosions, as many of its supercomputer forebears have done."

Sigh. But will it run Windows Vista?

[INRM]

PaxImperium,

But how would these computers render the scientific method obsolete? The information would still need to be obtained which could qualify as an observation. The information would need to be ordered to some extent for a hypothesis to be gauged, and the testing would need to be done by computer.

But all the steps of the scientific method are there...


Dr H,

"When it's installed at Lawrence Livermore National Laboratory in 2012, it could make new kinds of calculations possible

Let's hope that they don't start making all sorts of simulations which involve simulating sentient beings. It's bad enough that DARPA is working with IBM to create a computer simulation of a human brain to help make faster calculating computers that use less power and such things.


INRM

[erlando]

But how would these computers render the scientific method obsolete?

Where did it say anything about rendering the scientific method obsolete? How did you come to that conclusion?

Of course it is not going to render the scientific method obsolete. How could it? Computing power does not in any way imply AI or anything of the sort.

ETA:
Let's hope that they don't start making all sorts of simulations which involve simulating sentient beings. It's bad enough that DARPA is working with IBM to create a computer simulation of a human brain to help make faster calculating computers that use less power and such things.

Oh... Now I see..

Do you know anything about the structure and programming of computers? If you did you would know that today's computers are entirely deterministic in nature. They cannot in any way evolve sentience or anything like that. It is inherent in both the nature of the hardware and the programming.

Scientists studying the brain say that we understand very little about how it actually functions. Such understanding would be imperative for us to be able to simulate the human brain.

[INRM]

erlando,

Keep in mind DARPA actually did develop a cognitive computing contract which it gave to IBM. And it's goal was what I said it was.

Isn't the human brain deterministic? It's structure/genetics versus information that goes in?

[erlando]

Keep in mind DARPA actually did develop a cognitive computing contract which it gave to IBM. And it's goal was what I said it was.


So? A contract doesn't change the state of computing today.

We may never see sentient machines. There is nothing in computing theory that indicates that a programmed machine can exceed its programming. On the contrary very simple problems exists that a computer can never solve in finite time.

Isn't the human brain deterministic? It's structure/genetics versus information that goes in?

I wouldn't call the brain deterministic, no. For instance neurology describes an idea as a short-circuit - a connection between neurons that's not usually there. Something like that can never happen in hardware with positive results.

Also the human brain is capable of learning. Something computers are inherently bad at.

So contract or no contract - machine sentience is a very long way away and may never happen.

[Vorticity]

Since May 25, 2008. That's when Roadrunner at Los Alamos first acheived >= Petaflop speed. I was there, and have some code running on Roadrunner as I type this.

http://en.wikipedia.org/wiki/IBM_Roadrunner

[shadron]

"With 20 petaflops of computing power, meteorologists could predict local weather down to the 100-meter range. For an event like a tornado, that could mean being able to predict the path that the twister takes through a town, allowing for targeted evacuations that save lives. "

"When it's installed at Lawrence Livermore National Laboratory in 2012, it could make new kinds of calculations possible, but initially, that power will be primarily used to simulate nuclear explosions, as many of its supercomputer forebears have done."

Sigh.

Yeah, it might be able compute it. The problem is that you have to set the initial conditions first, and measurements like that (wind speed, direction, temperature, pressure, etc) in a 100 meter cubic lattice that includes a tornado or pre-tornado conditions are really hard to come by. It could generate data from some theoretically defined initial conditions, but the conclusions it comes to will only be as significant as the differences between the real and theoretical conditions, amplified by the pseudo-chaotic processes. What it will do is enable better understanding of those proccesses.

The NCAR lab in Boulder, CO is dedicated to doing this, only it does not limit itself to tornadic conditions.

[Dr H]

But will it run Windows Vista?

ROTFL!

Thanks, I needed that. :D

[Dr H]

Do you know anything about the structure and programming of computers? If you did you would know that today's computers are entirely deterministic in nature. They cannot in any way evolve sentience or anything like that. It is inherent in both the nature of the hardware and the programming.

Are you sure we know how sentience evolved?

Scientists studying the brain say that we understand very little about how it actually functions. Such understanding would be imperative for us to be able to simulate the human brain.

Unless, of course, they did it by accident. :eek:

[erlando]

Are you sure we know how sentience evolved?

No, but I'm pretty sure that our pre-sentient ancestors didn't have deterministic "brains" based on boolean logic.

Unless, of course, they did it by accident. :eek:

I'm still of the conviction that machine sentience is as near to impossible as you can get with the current way of structuring hardware and coding software.

"Accidents" in computer hardware or software leads to errors and crashes. Code is very VERY sensitive to errors. Even the most advanced neural nets crashes if "mutations" happen in the code.

Science understands very little about how the brain works and even less about how thought works. How are we supposed to be able to simulate thought if we don't know how it works?

[Wildy]

But will it run Windows Vista?

No. Nothing can run Vista.

[drainbread]

Since May 25, 2008. That's when Roadrunner at Los Alamos first acheived >= Petaflop speed. I was there, and have some code running on Roadrunner as I type this.

http://en.wikipedia.org/wiki/IBM_Roadrunner

http://en.wikipedia.org/wiki/Folding_@Home

Beat Roadrunner by over a year.

[Dr H]

No, but I'm pretty sure that our pre-sentient ancestors didn't have deterministic "brains" based on boolean logic.

I'm not sure how you're so sure about that, but I don't have any particular stake in the question one way or the other.

I'm still of the conviction that machine sentience is as near to impossible as you can get with the current way of structuring hardware and coding software.

"Accidents" in computer hardware or software leads to errors and crashes. Code is very VERY sensitive to errors. Even the most advanced neural nets crashes if "mutations" happen in the code.

Science understands very little about how the brain works and even less about how thought works. How are we supposed to be able to simulate thought if we don't know how it works?

Since we don't really have a solid undisputed definition for "sentience", it's true that we are probably not currently equipped to create it "artificially," as it were.

But are you familiar with the concepts of epiphenomenalism? It may turn out that sentience is merely a side-effect of having a certain number of processes in operation beyond a certain level of complexity.

[Soapy Sam]

Petaflop ability started for me when I was about 48.