Simulating Real World: What's Needed?

[aquaregia]

You've probably heard about the thing where scientists simulated 10 seconds of half of a mouses brain with a supercomputer. Does anyone know how fast of a computer it would take to simulate an ants brain?

On a related question, let's say you wanted to simulate 1 cubic milimeter of the real world. And you wanted to be able to simulate ANYTHING that can happen in that cubic milimeter (ants brain, chemical reactions, nuclear reaction) down to the lowest level describable by modern physics (quantum mechanics, or the standard model) in real time. How big of a supercomputer would you need to do that?

I think that there should be a way to convert all the different processing speed measures into some standard measure which was based on how large an area of space that that computer can simulate in real time. Or if you use FLOPS as the standard measure, how many FLOPS does say 1 atom have (looking at it from the point of view of how many FLOPS it would take to simulate that 1 atom)?

 

[mgb_phys]

You can't even write the wave equation for anything more complicated than a lone Hydrogen atom on it's own in space - so the electric field interactions in neurons at the atomic level is probably pushing it!

On the other hand ant's only have something like 50,000 neurons so you should be able to do a reasonable job of simulating it at that level.

 

[aquaregia]

"PetaVision models the human visual system—mimicking more than 1 billion visual neurons and trillions of synapses.

"On Saturday, Los Alamos researchers used PetaVision to model more than a billion visual neurons surpassing the scale of 1 quadrillion computations a second (a petaflop/s). On Monday scientists used PetaVision to reach a new computing performance record of 1.144 petaflop/s. The achievement throws open the door to eventually achieving human-like cognitive performance in electronic computers.

This was recently done with the Roadrunner supercomputer. The Roadrunner has something like 12,000 PS3 chips and 6,000 other chips in it. I think there are something like 10 million PS3s sold. There are at least 50 computers for every PS3, so just doing some quick math, it should be possible to hook up all those computers and PS3's to make something with 27,000 times the processing power of the Roadrunner.

If the Roadrunner can do 1 billion neurons, and the human brain has 100 billion neurons, that means it would be possilbe to make a computer simulation of a brain that had 270 times the complexity of a human brain, using just this small subset of the total computing power of the worlds computers.

 

[Borek]

That's all assuming you know structure of things simulated. At the moment we can throw in billion of randomly connected neurons - but that'll be GIGO (garbage in, garbage out).

Still, you are probably right that necessary power is almost there, we just have no idea how to properly use it.

The tiniest insects are able to feed, move and replicate, and they are quite effective at that. Their brain is so small that I would be not surprised if it can be simulated on my mobile. Yet automatic vacuum cleaners - with designated task much easier than that of insect - are still stupid enough to block themseves between chair legs. We just don't know how yet.