• Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.

Simulating one second of real brain activity : 40 minutes, 83K processors, 1PB of RAM

Status
Not open for further replies.
http://gigaom.com/2013/08/02/simula...ain-activity-takes-40-minutes-83k-processors/

A team of Japanese and German researchers have carried out the largest-ever simulation of neural activity in the human brain, and the numbers are both amazing and humbling.

The hardware necessary to simulate the activity of 1.73 billion nerve cells connected by 10.4 trillion synapses (just 1 percent of a brain’s total neural network) for 1 biological second: 82,944 processors on the K supercomputer and 1 petabyte of memory (24 bytes per synapse). That 1 second of biological time took 40 minutes, on one of the world’s most-powerful systems, to compute.

If computing time scales linearly with the size of the network (a big if; I have no idea if this would be the case), it would take nearly two and half days to simulate 1 second of activity for an entire brain.

Nothing groundbreaking came from this research, but this just puts into perspective how insanely complex our brains are. Go science!
 

The Technomancer

card-carrying scientician
I feel like brain simulation and artificial intelligence is going to require a radically, fundamentally different computing architecture.
 

Sub_Level

wants to fuck an Asian grill.
tumblr_mny692yLfR1rs1wvso1_500.gif
 

luoapp

Member
K-supercomputer may not be the best choice for the task. 82,944 cores sound impressive, but it can fit on 31Tesla cards , just saying.
 

Sub_Level

wants to fuck an Asian grill.
Well, I can talk to myself in my head. And nobody can hear it.

It better be complex in there.

Not only can you talk to yourself in your head, you can think of yourself talking to yourself in your head while thinking of yourself talking to yourself in your head while thinking of yourself talking to yourself in your head.
 

Kainazzo

Member
So if Moore's Law stays in effect, then in just 9 years a computer will be able to simulate 1 second of the entire brain in a little under an hour?

In 18 years they'll manage 1 second in slightly under a minute.
In 30, computers should be able to process our brains 5 times faster than we can.

Or did I totally mess that up?
 
Kurzweil am cry. That motherfucker is definitely not going to make it to the singularity. He should turn his vitamin addiction into a nice healthy drug addiction.

ray2.jpg


He's not a scientist, let alone understands anything about neurology . People need to stop treating him as such.
 
So if Moore's Law stays in effect, then in just 9 years a computer will be able to simulate 1 second of the entire brain in a little under an hour?

In 18 years they'll manage 1 second in slightly under a minute.
In 30, computers should be able to process our brains 5 times faster than we can.

Or did I totally mess that up?


Moores Law actually says nothing about capability. It just says every 18 months or so you should be able to double the transistor count. Typically, it goes hand in hand with better performance, but it's not a guarantee.

Note, that time frame actually came from an Intel executive and is starting not to match up with reality.
 

industrian

will gently cradle you as time slowly ticks away.
Simulating a brain's thought process?

So they spent 40 minutes, 83K processors, 1PB of RAM just to come up with:

return "sex";
 

jchap

Member
If we can simulate a human brain we can make computers worse as chess and humans will have a chance again!
 

tokkun

Member
So if Moore's Law stays in effect, then in just 9 years a computer will be able to simulate 1 second of the entire brain in a little under an hour?

In 18 years they'll manage 1 second in slightly under a minute.
In 30, computers should be able to process our brains 5 times faster than we can.

Or did I totally mess that up?

Many chip companies are developing chips that are specifically designed for brain-like computations. I imagine that these will be much more efficient than the existing simulators, possibly by orders of magnitude.
 
Moores Law actually says nothing about capability. It just says every 18 months or so you should be able to double the transistor count. Typically, it goes hand in hand with better performance, but it's not a guarantee.

Note, that time frame actually came from an Intel executive and is starting not to match up with reality.

Yup. And also the brain is all about interaction and bio-chemical processes which is going to take decades to figure out even with computers and simulators that may well exceed the "speed" of the brain.

Also, Kurzweil's explanation of exponential knowledge, especially in biology is arbitrary. He made it up to fit his graphs.
 
Can this be a strong AI vs. weak AI proponents thread please? I could do with a good chuckle.

Also: Moore's law is slowing down, so the prognosticators might want to adjust their calculations to a smaller base for the exponential.
 
SO MUCH less than I would have expected.

I have incredible optimism that there will be some sort of insane breakthrough within my lifetime.
 
So if Moore's Law stays in effect, then in just 9 years a computer will be able to simulate 1 second of the entire brain in a little under an hour?

In 18 years they'll manage 1 second in slightly under a minute.
In 30, computers should be able to process our brains 5 times faster than we can.

Or did I totally mess that up?

As mentioned, Moore's Law is slowing down., we're gonna need to find a better architecture to do this with...
 

DopeyFish

Not bitter, just unsweetened
Question is... How much of that activity is dedicated towards consciousness vs sensory and general system functions?


Agreed. When I was learning about computer architecture it was something that was painfully obvious. I miss that class.

Depends on approach. Brute forcing (synapse simulation), yes.

AI is probably possible today but problem is knowing how to construct a consciousness. We aren't anywhere close to that, I think.
 
I don't know if I'm stupid or something, but I have a lot of questions about the (future) accomplishments and morality of this.
A fully functional (human) brain doesn't just spring into existence. To tackle that, have they simulated the whole development of the system from the beginning (which would also require simulating some kind of surroundings and other organisms)? If not, what kind of preset neural network are they simulating? Is it truthful?
In the end, our brains are just machines, too. In terms of "soul" and awareness, any accurate simulation would be equal to us humans. Couldn't these kinds experiments be a torture to an aware being?
 
I feel like brain simulation and artificial intelligence is going to require a radically, fundamentally different computing architecture.

Massively parallel networks like this are one of a couple of practical areas of computing that quantum computing could seriously improve. But we're decades away from building a machine capable of doing such a simulation.
 
I don't know if I'm stupid or something, but I have a lot of questions about the (future) accomplishments and morality of this.
A fully functional (human) brain doesn't just spring into existence. To tackle that, have they simulated the whole development of the system from the beginning (which would also require simulating some kind of surroundings and other organisms)? If not, what kind of preset neural network are they simulating? Is it truthful?
In the end, our brains are just machines, too. In terms of "soul" and awareness, any accurate simulation would be equal to us humans. Couldn't these kinds experiments be a torture to an aware being?

We are very, very far away from creating a conscious machine. I wouldn't start worrying about that just yet.
 
In the end, our brains are just machines, too. In terms of "soul" and awareness, any accurate simulation would be equal to us humans. Couldn't these kinds experiments be a torture to an aware being?
You're assuming quite a lot here. Most importantly, those views assume a purely naturalist premise as well as side-stepping the hard problem of consciousness.
 
Status
Not open for further replies.
Top Bottom