IBM cracks open a new era of computing with brain-like chip

[Fulgrim]

 

Scientists at IBM Research have created by far the most advanced neuromorphic (brain-like) computer chip to date. The chip, called TrueNorth, consists of 1 million programmable neurons and 256 million programmable synapses across 4096 individual neurosynaptic cores. Built on Samsung’s 28nm process and with a monstrous transistor count of 5.4 billion, this is one of the largest and most advanced computer chips ever made. Perhaps most importantly, though, TrueNorth is incredibly efficient: The chip consumes just 72 milliwatts at max load, which equates to around 400 billion synaptic operations per second per watt — or about 176,000 times more efficient than a modern CPU running the same brain-like workload, or 769 times more efficient than other state-of-the-art neuromorphic approaches. Yes, IBM is now a big step closer to building a brain on a chip.

 

The animal brain (which includes the human brain, of course), as you may have heard before, is by far the most efficient computer in the known universe. As you can see in the graph below, the human brain has a “clock speed” (neuron firing speed) measured in tens of hertz, and a total power consumption of around 20 watts. A modern silicon chip, despite having features that are almost on the same tiny scale as biological neurons and synapses, can consume thousands or millions times more energy to perform the same task as a human brain. As we move towards more advanced areas of computing, such as artificial general intelligence and big data analysis — areas that IBM just happens to be deeply involved with — it would really help if we had a silicon chip that was capable of brain-like efficiency.

 

Source

[MattP123]

Cool

[killers993]

Can it run Crysis? 

[Thorium19]

Neat, that's some incredible performance from it, but it depends what application they use it for. Now all we need is to create the six million dollar man. Or Jarvis. I vote Jarvis, so long as I can have an iron man suit.

[AMICLG]

Neat, that's some incredible performance from it, but it depends what application they use it for. Now all we need is to create the six million dollar man. Or Jarvis. I vote Jarvis, so long as I can have an iron man suit.

It's to create a deep neural network, I doesn't function like a normal computer.

[qwertywarrior]

so its as smart as a cockroach but still that great

early baby steps

 

anyways if they overclocked it

it would save them for making larger chips ,no ?

[Linus_torvalds]

It has 28 nm 5.4 bilion transistors. how the hell it consumes 12451326 less power then my 22 nm i7 which has 1.4 billion transistors

[APrettyCoolWalrus]

I have to overclock my brain brb

[patrickjp93]

It has 28 nm 5.4 bilion transistors. how the hell it consumes 12451326 less power then my 22 nm i7 which has 1.4 billion transistors

Low clock rate and severely reduced instruction set. No overclocking, no internal execution security. There's no bloat hardware.

[bob345]

This is pretty awesome

[Spongy141]

So our brains are clocked at 20 EHz...

[Tech_Dreamer]

Sounds cool , But here's something experts have to say, Human brain is far more advanced than a processor chip it doesn't think in binary , it's way too complex for even the best scientist to understand hard to replicate as it evolves & changes, not just a stagnant byproduct from a factory , altho this might process things on an analytical pattern , taking every probability into account & calculating them & shortening the processing time by having faster clock speeds or processing power it might replicate quicker response, But, nothing can be matched to the intelligence of human brain's way thinking , it's complex & versatile it might sound similar but it's just a huge a processing chip at the end nevertheless,

 

Calling it brain is somewhat far fetched , even if AI is present , it goes through same pattern,

[Mooshi]

But can it overclock and be relevant to us?

[patrickjp93]

So our brains are clocked at 20 EHz...

Reread the page. We perform at about 200 Exa Flops, but our neurons only fire at 200 Hz max.

[Spongy141]

Reread the page. We perform at about 200 Exa Flops, but our neurons only fire at 200 Hz max.

200 Hz max but they only are at 4 Hz (Normal everyday functions) through 13 Hz (Deep sleep), though if we could overclock them to be at even 50 Hz, or that 200 Hz max; just imagine how much smarter you'll be.

[Sharif]

Future!~

very exciting 

[Syntaxvgm]

I wonder how this will affect, if at all, the simulations of the brain? The most recent and successful project simulated something like 1% of 1 second of brain activity in 40 minutes, so maybe this will bring us closer? 

[Kingofpants]

200 Hz max but they only are at 4 Hz (Normal everyday functions) through 13 Hz (Deep sleep), though if we could overclock them to be at even 50 Hz, or that 200 Hz max; just imagine how much smarter you'll be.

You probably would do maths faster but I wouldn't say you would be smarter in the sense

*You have more Memory

*Your Pathways are more efficient

If anything you would probably have massive headaches all the time due to the amount of energy your brain would be consuming.

But you probably would

*Do maths faster

*Think of solutions faster

*"Read" memory faster

*Have very little motion blur

But ideally you would get More neurons, more effecient pathways (They already are fairly effecient ) and have Faster neuron connections where its 2500m/s + rather than 150 or so m/s

[PortentousLad]

Can it run Crysis?

Probably not, but that's to do with OS support

[Sidiox]

Sounds cool , But here's something experts have to say, Human brain is far more advanced than a processor chip it doesn't think in binary , it's way too complex for even the best scientist to understand hard to replicate as it evolves & changes, not just a stagnant byproduct from a factory , altho this might process things on an analytical pattern , taking every probability into account & calculating them & shortening the processing time by having faster clock speeds or processing power it might replicate quicker response, But, nothing can be matched to the intelligence of human brain's way thinking , it's complex & versatile it might sound similar but it's just a huge a processing chip at the end nevertheless,

 

Calling it brain is somewhat far fetched , even if AI is present , it goes through same pattern,

Actually, our brains do run in a sort of binary way. An impuls can either go or not go, but this 1/0 state can be affected by multiple other synapses. Very binary like, but not very computer like. 

 

Anyway it is cool and also somewhat scary that we are coming close to approaching the brain, because I think that on these chips you could also run low level AI's, since well you created a brain. I wonder how these machines are going to get used.

[SirPaul]

Sounds cool , But here's something experts have to say, Human brain is far more advanced than a processor chip it doesn't think in binary , it's way too complex for even the best scientist to understand hard to replicate as it evolves & changes, not just a stagnant byproduct from a factory , altho this might process things on an analytical pattern , taking every probability into account & calculating them & shortening the processing time by having faster clock speeds or processing power it might replicate quicker response, But, nothing can be matched to the intelligence of human brain's way thinking , it's complex & versatile it might sound similar but it's just a huge a processing chip at the end nevertheless,

 

Calling it brain is somewhat far fetched , even if AI is present , it goes through same pattern,

 

I don't know you so i will be kind to you

Just google neural network AI for example, and try to imagine .  We cannot recreate the brain , but we can mirror it with calculations. 

AI in games is using the dumb AI system based on behaviors, AI in robots kinda also, although few of them are using the neural network.

AI with neural network is really powerfull thing but you neeed many processing nodes (neural node) to process, for which this chip is build for.

Also neural network AI is mostly not used in robots/games but in pattern recognition and image processing , so basically with such huge power operating neural network this could be really great news for the cancer research e.c.t.

 

[mr moose]

Can it run Crysis? 

no, it will be completely shit house with mathematical functions and possibly not able to carry out more than one task/process at a time,   but it will be exceptional with lateral problem solving and tasks that require adaptive non-defined resolution. 

[patrickjp93]

I wonder how this will affect, if at all, the simulations of the brain? The most recent and successful project simulated something like 1% of 1 second of brain activity in 40 minutes, so maybe this will bring us closer?

With Google Brain we simulated a full bee brain for several hours. Now with 2 Titan Z you can do the same thing.

[mr moose]

With Google Brain we simulated a full bee brain for several hours. Now with 2 Titan Z you can do the same thing.

I knew the titan z's were good for something,  no one believed me but I said they'd be good for something... :blink:

[Askew]

Wow, I thought you'd written a really good article there until I clicked the source and saw that it was copy/pasted. :lol: