China may become the first country to turn on a 100 petaflops supercomputer, just one order of magnitude away from "exascale":
A little over one year ago, export blocks put in place by the US government threatened to derail China's plans to upgrade its Tianhe-2 supercomputer, the world's fastest since June 2013, to its originally planned peak capacity of 100 petaflops. At the time, many in the industry anticipated that the efforts to block China's supercomputing capability by banning access to US technology from Intel and other hardware vendors would backfire.
Indeed, China was sufficiently incentivized to redouble efforts on its homegrown supercomputing effort and it had the cash from the squashed Intel deal to do it. A couple months after news of the blacklist came out, China revealed plans to build not one, but two 100-petaflops supercomputers using a variety of native chip, accelerator and interconnect technologies. One of these systems was a fully-realized Tianhe-2, which was slated for a late-2016 launch.
VR World, the same publication who broke the blacklisting story last year, is now reporting that China is on track to announce a 100-petaflops supercomputer in June, during the 2016 International Supercomputing Conference in Frankfurt, Germany. China had originally said it would have such a system in late-2016, but this is the same country that launched its 33-petaflops (LINPACK) Tianhe-2 two years early.
Related Stories
New information has emerged about China's exascale plans, which are a part of China's 13th five-year plan for 2016-2020. Despite U.S.-imposed export restrictions on processors, two 100 petaflops systems will be launched sometime during 2016, possibly as soon as the 2016 International Supercomputing Conference in June. One of these systems will be an upgrade to Tianhe-2, and both may utilize homegrown accelerators.
At least one exascale prototype system will be built prior to a 1 exaflops system:
The exascale prototype will be about 512 nodes, offering 5-10 teraflops-per-node, 10-20 Gflops/watt, point to point bandwidth greater than 200 Gbps. MPI latency should be less than 1.5 µs, said [Beihang University Professor Depei] Qian. Development will also include system software and three typical applications that will be used to verify effectiveness. From there, work will begin on an efficient computing node and a scheme for high-performance processor/accelerator design.
"Based on those key technology developments, we will finally build the exascale system," said Qian. "Our goal is not so ambitious – it is to have exaflops in peak. We are looking for a LINPACK efficiency of greater than 60 percent. Memory is rather limited, about 10 petabytes, with exabyte levels of storage. We don't think we can reach the 20 megawatts system goal in less than five years so our goal is about 35 megawatts for the system; that means 30 Gflops/watt energy efficiency. The expected interconnect performance is greater than 500 Gbps."
(Score: 2) by bob_super on Friday April 15 2016, @05:20PM
Don't worry, we're still the best at filling databases.
(Score: 1, Insightful) by Anonymous Coward on Friday April 15 2016, @05:49PM
A little over one year ago, export blocks put in place by the US government threatened to derail China's plans to upgrade its Tianhe-2 supercomputer, the world's fastest since June 2013, to its originally planned peak capacity of 100 petaflops.
I wonder how the US would react when China puts in place export blocks on computer chips or other hardware because "it could be used to do computations in order to build nuclear weapons"... (heck, what about an Export Ban on iPhones... that would really cause US citizens to call their congress-critter)
My point is this: when the US puts in place export controls, everyone goes "oh, that makes sense, yeah, sure, go ahead" but when someone else does it suddenly *they* are the biggest protectionist, communist or nationalistic scum on earth.
I say: Well done China, on becoming more self-sufficient.
(Score: 3, Insightful) by jcross on Friday April 15 2016, @06:26PM
Yeah anyone could have and many did guess how this would turn out. One of the few things the USA still dominates at, and we tell the largest manufacturing economy in the world "no don't buy it from us, make it yourself" as if they needed any encouragement to get up in our rice bowl.
(Score: -1, Troll) by Anonymous Coward on Friday April 15 2016, @06:32PM
You know what I'm tired of? Heterosexual males raping and sodomizing young girls. That's what I'm tired of.
(Score: 0) by Anonymous Coward on Friday April 15 2016, @06:03PM
I believe that like I believe penis extension enhancing drugs.
(Score: 2) by takyon on Friday April 15 2016, @06:33PM
Their top system is already 33.8 petaflops RMAX / 54.9 petaflops RPEAK. What basis do you have to doubt this news? Or are you just going to rail against LINPACK?
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 3, Informative) by bitstream on Friday April 15 2016, @06:15PM
Actually the estimated capacity to simulate a human brain is 36.8 * 10^15 flops and 3.2 * 10^15 bytes storage to go with it.* [hplusmagazine.com].
So now, when will we see the supercomputer actually talking to us?
Some obstacles may happen on the way though. Some estimate that the required capacity is 10^12 times or so larger. This will take more time to accomplish, mainly due to that neurons are more complex than thought to simulate. The next problem may be if neurons require quantum computing to actually work, which these computers don't support (asfaik).
(Score: 3, Interesting) by takyon on Friday April 15 2016, @06:36PM
Using a neuromorphic [soylentnews.org] system is a far better idea.
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 3, Interesting) by devlux on Friday April 15 2016, @08:30PM
"Researchers estimate that it would require at least a machine with a computational capacity of 36.8 petaflops (a petaflop is a thousand trillion floating point operations per second) and a memory capacity of 3.2 petabytes – a scale that supercomputer technology isn’t expected to hit for at least three years."
All of this ignores the fact that running a silicon based digital computation of an analog biological system is a horribly inefficient way of going about this.
The human brain is highly plastic, meaning you can survive and even thrive after serious traumatic brain injury in many cases with little or no change to personality.
This means that large sections of the brain are fully redundant. However even a small stroke in the wrong area of the brain can leave you locked in and unable to function anymore, these sections are clearly not redundant. Age too is a factor, as we age the likelyhood of the brain being able to rewire itself after trauma quickly begins to taper off.
We don't need to simulate an entire brain. For example, the motor functions are basically irrelevant, as are the autonomic functions. That's 20% gone right there.
Visual processing is already a solved problem, audio is getting there. Touch is being worked on, and smell is just chemical analysis of the air.
The synthesis of sensory input along with short and long term memory are the nuts left to crack.
But I have doubts we're going to brute force them in any meaningful way.
Yet we are merging with our machines even now.
How many times a year do people die from GPS related mixups?
Do you have a kid under 25? Ever try to remove their arm err cellphone? It may as well be a surgical procedure.
These machines are intimate with us, they sense us and they grow more and more capable in terms of computation, sensation and recall.
As they get smaller and more capable they will become a natural part of us. Prosthetics are the first step, enhancements are a likely next step.
What happens when we get general purpose nanotech that functions as a programmable "immune system" enhancement?
A single shot at birth, that can be reprogrammed on the fly to detect and eradicate illness, and repair diseased tissue?
How long do we actually live beyond that point? When does the organic you cease to be, replaced cell by cell with something more robust and more flexible?
I'm not talking about a "grey goo" scenario here, but basically the opposite.
A body that stays eternally in it's prime, no telomere expiration date.
At what point do we become our own Theseus paradox ? https://en.wikipedia.org/wiki/Ship_of_Theseus [wikipedia.org]
When do we stop saying "We are all human" and start being something much more?
That's how I believe that generalized AI will come to be. Machines which stay with us our whole lives, eventually becoming the information carrier for "US". While we stay the same and yet become so much more.
(Score: 3, Insightful) by bitstream on Friday April 15 2016, @09:42PM
If one can get brain functionality with way less neurons then we should be way past the simulation point by now. The question then becomes when the software will be functioning. Any ideas?
Perhaps the Theseus paradox fools the reader a bit. If something has their functions replaced with the same functions. It is functionally the same. The catch comes when you have stuff like quantum mechanics etc where every dimensional cross point is unique and will not be repeated. So as long as the replacements are functionally equivalent it shouldn't matter. If a PC is implemented in FPGA, is it still a PC? I would say yes because it's deterministic. But once you have anything that is pathway dependent and is affected by environmental factors, it's in essence a random walk and never the same even if the starting point is a perfect copy. Two flowers with the same DNA and phenotype will have different outcomes despite a very similar start.
(Score: 2) by devlux on Friday April 15 2016, @10:56PM
Your body renews & replaces nearly every single cell about every 7 to 15 years.
https://www.youtube.com/watch?v=XqJWSyUbmkw [youtube.com]
Under your supposition, you are not the same person from 7 years ago and yet bill collectors would beg to differ.
You have a persistence of memory, a continuum of the experience of you being you (unless you don't in which case, seek immediate medical attention).
You believe yourself to be you. Ergo to you, you are you.
To the outside world you are also you, unless you're operating under a delusion of being someone else, in which case see my above comment about medical attention :)
To those who know you. You are still you, but this is a wholly different you than the canonical, definitive you, that you believe yourself to be.
It is a model of "you" that the others keep in their minds. This model is derived from what they know about you.
To us, you are the memory of our experiences and interactions with you.
My solution to the Theseus paradox.
You are nothing less and nothing more than the information that is you.
Information which has been fighting a never ending battle against entropy since before the first life forms emerged.
Everything else about you, is the form factor of the information carrier. The medium that allows the information to travel, mix, copy, increase in number and spread.
The carrier can change at any time and as long as the information that is you remains the same, then you remain the same.
Specifically addressing Theseus's ship.
It ceased being Theseus's ship the day Theseus ceased being aboard it.
The ship served as a carrier for Theseus, and it was the desire by others to keep Theseus's memory intact, which provided the onus for others to keep the ship intact in memoriam.
In other words, it was the memory of Theseus that drove others to keep the ship in good order and repair, but it had not been Theseus's ship in a very long time.
(Score: 3, Insightful) by bitstream on Saturday April 16 2016, @02:09AM
I may not be the same person, but I'm similar enough to pass as the same person. Everything changes, just not enough to be casually observed.
The other that you state means that if one can extract the memories, one can also duplicate the person good enough for a simulation at least.
(Score: 2) by devlux on Saturday April 16 2016, @04:06AM
Yes exactly. Sort of. It's not JUST your memories, it's also how you synthesize them along with how you collate them given new sensory input.
The combined function of the previous information acting on new information whatever that information may be.
The question asked was "if we're past the simulation point from a technological standpoint then how long will it be until we are able to simulate a human mind in a computer?"
My answer to that is, "I believe we will never be able to simulate a human mind with current approaches because they focus on the carrier rather than the information being carried."
Once we sort out how to map the information itself to new machinery, then the simulation pathway becomes clear.
Similar to how we moved from film to digital for movies. Or mechanical computation devices and techniques to electronic computational techniques.
There was no electronic mimic of this, https://en.wikipedia.org/wiki/Curta [wikipedia.org] prior to the advent of fully electronic computers and electronic computational techniques are very different from mechanical devices that do the same thing. It turns out the math, it's input and it's output, was the thing that mattered and not the machine performing the computations.
However it now begs the question, at which point are you no longer yourself and suddenly just a very complex machine simulating you.
It's much more philosophical and meta-physical than valid hard science. It's not a theory, it doesn't predict anything. It's just a philosophy that I use to frame these questions because it feels like we're on the cusp of something new and different.
Yet I can't help but feeling like these are questions we will need to answer eventually.
Only time will prove me right or wrong, but I like taking about it because, before we can truly define artificial intelligence, we need to find a way to define intelligence.
An information driven approach might provide us something quantifiable and objective rather than the nearly completely subjective measures used today.
Although it's PopSci, here is something fascinating to watch and think about, there is some hard science backing it.
https://www.youtube.com/watch?v=prfxjrz0_0k [youtube.com]
Time from 2:22 is particularly relevant to our discussion.
(Score: 2) by bitstream on Saturday April 16 2016, @08:12AM
There is no simulation just interpretations and similarities in varying degrees. Two perfect copies will take on a unique future. So copying will not make objects equal.
I seen documentaries about a neurosurgeon have an idea that our mind is a quantum computer at room temperature. He holds talks and writes papers on this. If it's quantum, then a fast computer isn't enough. I suspect in addition self feedback and random input is also part of what makes it work.
(Score: 2) by devlux on Saturday April 16 2016, @08:48AM
You're not alone in the "brain as quantum computer" camp.
See my response to toygeek over here... https://soylentnews.org/article.pl?sid=16/04/16/0132245 [soylentnews.org]
Would love to see a link to your neurosurgeon though. I like to keep an open mind. But the question isn't "is this or is this not a quantum computer", because everything is a quantum computer. It's "to what effects do the non-intuitive quirks of quantum mechanics dominate the relevant computations which define us".
(Score: 2) by bitstream on Saturday April 16 2016, @04:24PM
Seems my memory was wrong. He's an Anesthetist. Here's some articles on him:
Stuart Hameroff, MD consciousness and microtubules in a quantum world. Interview by Bonnie Horrigan. [nih.gov], and Anesthetics act in quantum channels in brain microtubules to prevent consciousness. [nih.gov].
Search for: Stuart Hameroff, quantum, consciousness, microtubules
And you find the motherload..
I think he might be on to parts of the puzzle to create a thinking machine for real (AI). And a clue to room temperature quantum computing. Even plants succeeds with harnessing quantum processes without super labs. And it can be proved the the efficiency by with sun energy is absorbed and used.
(Score: 5, Funny) by Dunbal on Friday April 15 2016, @06:35PM
And here I was thinking this was a new computing method that involved throwing animal rights activists from tall buildings.
(Score: 2) by devlux on Friday April 15 2016, @09:00PM
Ok I am sad to admit, the above joke was a walkaway.
I didn't get it. Walked away from the computer, down the street to the bus stop. The punchline did not hit for me until I got all the way to the bus stop.
That's a long drop.
Modded funny.
(Score: 0) by Anonymous Coward on Friday April 15 2016, @10:49PM
FTFY. Are you a member of PETA?
(Score: 2) by devlux on Friday April 15 2016, @11:00PM
Yes as a matter of fact I'm a card carrying member of PETA.
People for the Eating of Tasty Animals
(Score: 2) by Dunbal on Friday April 15 2016, @10:59PM
I'm sorry. My profession is known to attract those with a very dark sense of humor :)
(Score: 0) by Anonymous Coward on Friday April 15 2016, @11:50PM
You need to upgrade your humorous processor. It's lacking petaflops.
(Score: 0) by Anonymous Coward on Saturday April 16 2016, @05:36AM
Lacking petaflops? Watch some porn!
(Score: 2) by rts008 on Friday April 15 2016, @09:29PM
Sadly you were mistaken, but I will say the comment was indeed worthy of accolades!
Well done, and a tip of the hat to you... :-)
(Score: 2) by devlux on Friday April 15 2016, @11:34PM
Agreed which is why I modded it funny.
Just took a long time for the punchline to hit. That worries me, I like to consider myself quick witted. Normally I would have gotten that one off the bat.
(Score: 0) by Anonymous Coward on Saturday April 16 2016, @01:38AM
I hope they can get that thing up and running quickly so they can figure out the
predicted radiation contamination areas and evacuation plans after the nuke that powers it blows up.
Needs a PhD to know!