New IBM brain-like chip: 4096 cores, 1 million neurons, 5.4 billion transistors

[superunknown98]

Besides being event driven and having special programing tools how is this different from any other non-general purpose parallel architecture? Say a GPU or intel's Xeon Phi? I know those architectures are designed for graphics or simple calculations, but isn't IBM's new chip just a much larger manifestation of the same theory? Just having lot's and lot's of small simple cores all interconnected makes a brain chip?

 

[Burpo]

"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"

 

[witeken]

@superunknown98: Here is the core, much different from a regular core:

(1core = 1.3M transistors)

This chip isn't even based on the Von Neumann architecture. It also has synapses.

 

[Maximilian]

this is gpu or cpu ?

Its a neural net processor... a learning computer.

 

[VirtualLarry]

I'm not afraid of computer intelligence because that's real intelligence, not the evolutionary biological pseudo intelligence of humans.

I think that's very, very wrong, and a very disturbing thought / viewpoint.

Intelligence (and most importantly, free will) is a human trait. Computers can never replace that (though they might surpass "intelligence" in some metrics, arguably benchmarked according to computational ability).

If computers become "smarter", will they create Art? Will they believe in God?

 

[jhu]

Intelligence (and most importantly, free will) is a human trait. Computers can never replace that (though they might surpass "intelligence" in some metrics, arguably benchmarked according to computational ability).

If computers become "smarter", will they create Art? Will they believe in God?

Actually, this is the wrong point. Intelligence is not just a human trait. There are plenty of non-human animals that are intelligent. As for free will, it's a nice illusion that doesn't exist.

 

[SAAA]

I always wondered if we were simply pushing too much on the standard CPU/GPU logic, especially I asked myself why the hell do we need a billion transistors working a billions time a second each one to do maybe some billions of simple computation, in binary... There must be a smarter way to use all those expensive switches no?

Hence I'm both surprised by this device and totally fascinated by the idea of a future scaling of intelligence more than hertz or any other meaningless number.
Oh just get us that singularity, wont you IBM?

To anyone worried enough about its existence just remember how dangerous are ourselves humans, to the skeptics that think it's impossible to achieve intelligence just look at us: what's more than a dozen of atom kinds placed in the right way in our brains?
If simulation at the atom level is almost impossible then individual neurons are much more easy and this device already tries so.

Intelligence (and most importantly, free will) is a human trait. Computers can never replace that (though they might surpass "intelligence" in some metrics, arguably benchmarked according to computational ability).

If computers become "smarter", will they create Art? Will they believe in God?

1) almost sure.
2) hmm, ask them. Unless we start to believe them so...

Actually, this is the wrong point. Intelligence is not just a human trait. There are plenty of non-human animals that are intelligent. As for free will, it's a nice illusion that doesn't exist.

Haha nice one, 50% agree here: free will? That's maybe an illusion but it depends on how really the world works, say I need a time machine to discover it. Paradoxes allowed? Then possibly yes. Fixed time events? Then it's no-no. Parrallel universes aside of course

 

[VirtualLarry]

Actually, this is the wrong point. Intelligence is not just a human trait. There are plenty of non-human animals that are intelligent.

By what metric? That they can add simple numbers and stomp their hooves?

Face it, humans are the only species to master technology. ("Tools to make tools" - to use a programmer's expression.)

As for free will, it's a nice illusion that doesn't exist.

I strongly disagree with that.

 

[jhu]

By what metric? That they can add simple numbers and stomp their hooves?

Face it, humans are the only species to master technology. ("Tools to make tools" - to use a programmer's expression.)

Wrong. Many animals are able to use tools; a few also make them.

I strongly disagree with that.

Then you must accept that protons, neutrons, and electrons have some degree of free will because that's what we're made of.

 

[VirtualLarry]

Wrong. Many animals are able to use tools; a few also make them.

Note that I said "tools to make tools". Some mammals may make use of tools, but do they make tools to make tools? The distinction is important.

Then you must accept that protons, neutrons, and electrons have some degree of free will because that's what we're made of.

I suggest you read as many philosophy books as you read physics texts. It's an open question of whether our actual consciousness is actually outside or above our physical world.

 

[jhu]

I suggest you read as many philosophy books as you read physics texts. It's an open question of whether our actual consciousness is actually outside or above our physical world.

I minored in philosophy, focusing on the philosophy of mind. The question of whether consciousness is outside of the physical world is moot and pondering it is essentially mental masturbation. But that's beside the point. The issue of consciousness and free will are two distinct and non-overlapping discussions.

Most philosophers who ascribe to the non-physicality of consciousness have glommed onto it as an epiphenomenon (ie, consciousness rides alongside the physical manifestation). Then there are the physicalists who reject epiphenomenon. Both parties essentially don't see "free will" playing any part in their positions.

 

[Haserath]

I think that's very, very wrong, and a very disturbing thought / viewpoint.

Intelligence (and most importantly, free will) is a human trait. Computers can never replace that (though they might surpass "intelligence" in some metrics, arguably benchmarked according to computational ability).

If computers become "smarter", will they create Art? Will they believe in God?

Free will is governed by the mind, which is essentially a biological computer.

I'd say the only thing that separates us from animals is our extensive use of language. Without that, we'd be the same as every other animal.

 

[Maxima1]

Note that I said "tools to make tools". Some mammals may make use of tools, but do they make tools to make tools? The distinction is important.

Koko the gorilla knew sign language (usage of over 600 signs and knowledge of more than a 1000)

http://www.youtube.com/watch?v=SNuZ4OE6vCk

Koko responds to sad movie

http://myfox8.com/2012/06/16/gorilla-responds-to-sad-movie-viral-video/

"Koko the Gorilla has watched “Tea with Mussolini” several times on TV. When the sad part comes, she turns away and starts signing: “Frown, sad, cry, bad, trouble, mother and Koko-love.”

Koko cries over the loss of a kitten that she named and wanted

http://www.youtube.com/watch?v=CYD6KZsOjxw
http://articles.latimes.com/1985-01-10/news/mn-9038_1_pet-kitten

"Koko the "talking" gorilla whimpered with grief and "discussed" the death of her pet kitten, struck and killed by a car, for several days after getting the bad news, her teachers say."

I strongly disagree with that.

Free will has been utterly refuted. They've even assigned percentages to the factors that make you like heredity and the external/internal environmental factors. Not much left for the soul to influence....

http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1015201/

“Although the response bias was clear and predictable, the subjects were unaware of its existence. It is possible to influence endogenous processes of movement preparation externally without disrupting the conscious perception of volition.”

In the experiment, right-handed people would normally choose to move their right hand 60% of the time, but when the right hemisphere was stimulated, they would instead choose their left hand 80% of the time, yet they thought they did so under their own volition....

http://www.pnas.org/content/early/2013/03/14/1212218110

Even four seconds before a subject was conscious of having made a “decision” to add or subtract, the decision could be predicted from fMRI with almost 60% accuracy, which is very significant for a random expectation. This experiment also wasn't just dealing with a motor action, too.

Blindsight

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

http://www.nature.com/news/2008/081222/full/news.2008.1328.html

Blind Man Walking

"They then persuaded TN to set his stick aside and walk down a corridor strewn with lab equipment. He was able to do so flawlessly, despite being unable to consciously see any of the obstacles. Head down and hands loose by his side, he twisted his body to slalom slowly but surely between a camera tripod and a swingbin, and neatly stepped around a random series of smaller items.

At first he was nervous," says de Gelder. "He said he wouldn't be able to do it because he was blind." The scientists broke into spontaneous cheers when he succeeded. The results are reported today in Current Biology1.

Transient Global Amnesia

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

"A person experiencing TGA typically has memory only of the past few minutes or less, and cannot retain new information beyond that period of time. One of its bizarre features is perseveration, in which the victim of an attack faithfully and methodically repeats statements or questions, complete with profoundly identical intonation and gestures "as if a fragment of a sound track is being repeatedly rerun."

They end up like a broken record once their short-term memory keeps resetting.

 

[Engineer]

 

[SKORPI0]

It has begun...

http://terminator.wikia.com/wiki/Neural_Net_CPU

 

[Ajay]

And we quickly descend into P&N territory....
(Thread Tools -> unsubscribe)

 

[witeken]

I think that's very, very wrong, and a very disturbing thought / viewpoint.

Fair enough. I've never seen a neural computer with tens of billions of neurons. Those things don't exist yet, I don't know how they work, so I can only make hypotheses, which could be wrong.

Intelligence (and most importantly, free will) is a human trait. Computers can never replace that (though they might surpass "intelligence" in some metrics, arguably benchmarked according to computational ability).

I disagree. Conventional computers are just calculators. But nothing says that they have to be. If you design a computer with a different architecture, like those people at IBM did, it's very well possible that computers gain the intelligence and skills like we have instead of just doing mathematical operations on data.

If computers become "smarter", will they create Art? Will they believe in God?

I think that's for a large part dependent on the software. The 2 examples you give are subjective things that humans invented. Computers don't understand subjective things like color or emotions or bias. You'd have to replicate such things in the software so you eventually create a human brain on the chip. If you simulate a human, sure it's possible it will believe in god, because that's what a lot of humans do and replicating a human was the goal.

 

[cytg111]

The second you put "free will" or "god" into the equation it pretty much stops being a science experiment and starts being .. something else. From an evolutionary standpoint, if I invent true AI, is it really artificial or just the next big step?
The brain is a future-predicting device, no way around it, and so is AI. The better we get at AI the further we can have it predict. The scary scenanrio is where it is way smarter on the short run yet doesnt have an endgame.

 

[witeken]

Face it, humans are the only species to master technology. ("Tools to make tools" - to use a programmer's expression.)

Because there can be only 1 intelligent species per planet.

I strongly disagree with that.

Everything you do it determined by your brains. You have free will (that means: you can choose everything you do yourself and you also have the ability to search for information about what's the best thing you should do), but ultimately it's an illusion because your brain is quite determined by the laws of nature.

 

[witeken]

I suggest you read as many philosophy books as you read physics texts. It's an open question of whether our actual consciousness is actually outside or above our physical world.

Absolutely false. The brain is purely physical. A brain is ultimately just a computer designed by and for evolution.

 

[witeken]

I'd say the only thing that separates us from animals is our extensive use of language. Without that, we'd be the same as every other animal.

And humans have a lot more neurons.

 

[cytg111]

I suggest you read as many philosophy books as you read physics texts. It's an open question of whether our actual consciousness is actually outside or above our physical world.

2000's years of philosophy roughly equals 10 years of hard science. Philosophy is akin to religion, somewhere u turn to in the absence of answers.

 

[cytg111]

And humans have a lot more neurons.

Numbers game isnt really justice here. It is more like a neurons/weight ratio that is important.. More mass to move around means more neurons to command it.
Also problary why chicks are smarter than dudes, adjusting for the weight difference girls got bigger brains. And more connections / synapses too ... (sry guys, we are loosing this fight, but some of you allready know that.)

 

[cytg111]

And humans have a lot more neurons.

edit : Im in love with that girl ..

edit 2: this link https://www.youtube.com/watch?v=_7_XH1CBzGw as provided by witeken is a gem. If you havent allready then dont deny yourself this treat. Go. Watch it. Now.

 

[witeken]

Numbers game isnt really justice here. It is more like a neurons/weight ratio that is important.. More mass to move around means more neurons to command it.
Also problary why chicks are smarter than dudes, adjusting for the weight difference girls got bigger brains. And more connections / synapses too ... (sry guys, we are loosing this fight, but some of you allready know that.)

Good point. Mentally I was comparing it to chimpanzees, so less weight and 3X more neurons (and those neurons are also used for the right function: increasing intelligence).