IBM has won a $4.9 million government grant from DARPA to begin the first phase of research on “cognitive computing”– essentially building computers that work like living brains. The new brain-like computers will aim to process vast amounts of data to solve problems without relying on specific programmed algorithms. Mark Dean, Vice President of IBM said, “The challenge is that computers today are very good at computing, but what we really need is a more efficient way of sifting through information” [International Herald Tribune].
The inside of computers already have the look of neural networks, a static road map of electronic circuits. But the brain actually works by constantly creating, breaking, and tweaking the synaptic connections between neurons. Although today’s computers may excel at complex challenges with clear rules, like chess, they fail at simple tasks that require strategy, sensation, perception, and learning, like finding misplaced keys. IBM will partner with five universities to develop new nano-scale circuitry that has the ability to shift depending on the signals that pass through them. Free from the constraints of explicitly programmed function, computers could gather together disparate information, weigh it based on experience, form memory independently and arguably begin to solve problems in a way that has so far been the preserve of what we call “thinking” [BBC].
Dharmendra Modha, lead scientist on the project, says that the time is right for such a cross-disciplinary project because three disparate pursuits are coming together in what he calls a “perfect storm” [BBC]. First, neuroscience has discovered the biological processes underlying simple brain function. Second, supercomputing has enabled technology to match the complexity of these brains. And third, nanotechnology has made it possible to imagine creating simulated synapses, the very element of the brain that enables it to work at speeds and efficiencies computer scientists have so far only dreamed of [CNET News].
Modha says scientists have already created computers that rival the mental abilities of rats, but the new “thinking” computers will aim for the intelligence of cats. When asked whether he foresees any potential dangers to re-creating the brain’s powers, Modha hedged, “The questions are so technically challenging and it is so engaging at this point to begin to put the puzzle together, and the whole part is still not completely visible… I think it is premature to worry about what if” [CNET News].
Related Content:
80beats: Computers Take the Turing Test for Intelligence, But Fall Short
DISCOVER: Big Blue to Build a Brain
Image: IBM
November 22nd, 2008 at 10:31 am
“I think it is premature to worry about what if”
Famous last words…as robot domination begins!
November 22nd, 2008 at 1:55 pm
Interesting. But if they know how it works, then why not just let normal computers simulate brains, instead of creating new hardware?
November 22nd, 2008 at 6:06 pm
The hardware doesn’t support the type of intelligent computational model they’re trying to create.
Numenta is already working on this kind of technology and has been for some time. They even have a working tool they’ve written for existing computer hardware, meant for developer use.
Read the book “On Intelligence,” if you haven’t yet. Great read, even for those of us who aren’t neuroscientists. Might teach you a little about why you think the way you think.
November 23rd, 2008 at 12:35 am
I can’t even imagine what kind of coding would run hardware like this.
Hit the nail on the head Modernrocko, as it says in the article, conventional computers do an excellent job of well defined, mathematical problems. A machine of this nature would not be an amazing math cruncher, but would however be able to approach problems our conventional computers don’t deal well with.
Instead of hardcoding an artificial intelligence, a machine of this sort could learn, and just ‘know’ the answer, more like how we function. It would be a much better candidate for ’simulating’ natural language.
As for the what if question, I would agree that its both premature, and misplaced. Whether or not this project succeeds, its a step in the right direction to learning more about our own cognition.
November 23rd, 2008 at 5:53 am
Can’t they just throw in some magic spices (Darwinian Evolutionary Algorithms) and hope a Boeing 747 comes to be? Very interesting that they are using all this futile and obsolete engineering principles to “design” a circuit driven brain with the capacity to think like – maybe a cat or a rat. Junkyard and tornado apparently created the first cell phone. This could save all that 4.9 million dollars which I need so dearly for the purchase of my Ferrari Enzo and semi-monthly oil changes. We need different ways of thinking, Darwin obviously by himself and with his IQ of 120 left enough for us so that we will be able to stop thinking and start doing, pro actively.
November 23rd, 2008 at 1:15 pm
HEY I KNOW WHY DODNT WE JUST CLONE SOME BRAINS IN SOME JARS AND HOOK THEM UP TO SOME COMPUTER INTERFACES AND LET THEM LEARN THE HISTORY OF MAN AND WOMENKIND, MAYBE THEN WE WILL HAVE AN END RESULT, WHATS THE DEPARTMENT OF DEFENSE NEED WITH A THINKING COMPUTER ANY WAY.
CAN YOU SAY CYBERNETICS, OR HOW ABOUT THEY HAVE THE BIONICS NOW ALL THEY NEED ARE SOME BRAINS TO PUT THEM TO USE, WHY DOSENT ANY ONE WHANT TO ADVANCE AND ATTEMPT TO PERFECT WHAT WE KNOW THAT ALLREADY WORKS SO WHELL, LEAVE THE REASONING TO US, AND LET OUR TOOLS DO THE COMPUTATIONAL STUFF.
I DONT KNOW ABOUT YOU PEOPLE BUT I DONT THINK I WILL EVER HAVE A USE FOR A MACHINE OR COMPUTER FOR THAT MATTER THAT IS SMARTER THAN MANKIND, IT JUST DOESN’T SEEM LIKE A VERY USEFULL ENDEAVOR TO ME.
SQUACK BACK, PEOPLE WITH MORE INTELLIGENCE PLEASE ENLIGHTEN ME AS TO WHY AND WHAT WE NEED ARTIFICIAL BRAINS FOR, AFTER ALL IM PERFECTLY HAPPY EXERCISING MY BRAIN ON A DAILY BASIS, THATS WHY I READ.
November 23rd, 2008 at 9:11 pm
WHAT IF computers became smarter than humans? Would they then look at us like we look at ‘cognitively inferior’ beings? WHAT IF you actually could define intelligence in a way that doesn’t involve a numerical scaling system and therefore would not need to be formalized for computational purposes and could be ‘taught’ instead of ‘programmed’?
‘What if’ was the start of many scientific fields and it is never to early to hypothesize.
The time to set boundaries is now and the time to revise those boundaries is all the time. Escpecially when one of those ‘what if’s’ has to do with the end of the human race.
November 23rd, 2008 at 9:27 pm
@MikeD: “Instead of hardcoding an artificial intelligence, a machine of this sort could learn, and just ‘know’ the answer, more like how we function.”
Yeah, thats the idea. But we are disturbed, violent, sadistic bastards that learn all of our habits (bad and good) from people teaching them to us. I can’t even get my cat to sit on command. What makes people think they will be able to control this type of learning in machines?
My ANN’s never learn in a predictable way. Just like humans, cats, dogs etc.
November 24th, 2008 at 9:12 am
Yes, how will these machines (that will hopefully replace humans very soon, wink wink) know the difference between good and evil? Whats stopping an AI of this magnitude to learn from the Virginia Tech shooting? It seems that if you want to teach machines something, teach them whats morally right before they start to “get ideas”.
November 24th, 2008 at 10:17 am
The creationolistic aspect of this type of research is a way to play God.
Do we really need cognitive computing that has the potential to ‘think’ & learn’ as we do, before we resolve evolutionary embedded ’survival at any cost’ instincts that has always plagued us as a species?
If we succeed, what will these cognitive computers think about the race that spawned them?
November 24th, 2008 at 5:45 pm
“If we succeed, what will these cognitive computers think about the race that spawned them?”
Yes, this is interesting. As a software developer myself I have to write lots of conditional logic (ie: basically if , and, or etc… statements) I believe the engineers of this magnitude of AI require trillions of if statements from the hardware up to the software that will be running these machines. But, then they make this statement: “I think it is premature to worry about what if”. So they don’t worry about the end result of what they are creating?, they only worry about the hardware and subsequent code. If they don’t know the answer to the question then they don’t really know what they are creating or perhaps, they are blindingly committing suicide.
November 24th, 2008 at 7:42 pm
You guys buy too much into the media without thinking realistically about the consequences… I’m sure CERN had you all heavily worried as well too?
November 24th, 2008 at 8:11 pm
Why do it? So the computer can learn from natural language, for example. Why do it in hardware? So you can run it 24/7 without hurting it, and so that you can pull the plug if it messes up. Why build it with nano-scale, analog components? So it will be fast to train, unlike digitally simulated artificial neural networks. Why not throw some genetic algorithms at the problem? Because genetic algorithms and genetic programming are still algorithmic methods; too narrow, specific, and slow when implemented in digital computers.
November 25th, 2008 at 12:09 pm
Well, damn. I was going to go onto a rant about how we’re all DOOMED, DOOMED I SAY!, but I appear to have been beat to the chase.
Ah, well. Time to go spam some other comment board, then.
March 27th, 2009 at 2:59 am
Very interesting article you got there!
I will not be surprised if there would come a day when computers are able to think like humans but with better accuracy. However, I do think they still would lack the capacity of man to imagine things. That would be a great challenge to the computer scientists as well.
July 3rd, 2009 at 4:07 am
I do not think the “what if” question is misplaced at all, a little premature, yes. But if ever we will be successful in this endeavor, the danger will materialize very quickly and it will be rather late then to come up with rules and protocols. I think personally it’s a big dilemna; I’m so fascinated with the possiblity of human-made intelligence that rivals ours, yet at the same time apprehensive about the prospect of something that we creat will one day gain awareness of its own and realize it can be free from our governance.
August 17th, 2009 at 8:11 pm
I for one welcome our cat-level overlords!
Quick, distract it with the laser pointer!!