FACETS. Making Computers Work Like Brains

How can you get a computer to solve problems like a brain? How does the brain even solve problems to begin with? Below is an indepth look into FACETS, a groundbreaking consortium of European scientists and engineers that are taking long strides to bridging the gap between your brain and your PC. Give them enough time and money, and these cutting edge researchers may have a computer that doesn’t just act like a human…it will think like a human, too.

What is FACETS? What is it doing?

Fast Analog Computing with Emergent Transient States — with a daunting acronym that only a scientist could love, FACETS is a group of over 75 researchers with more than 10.5 million Euros to spend on tackling one major issue: finding out how the brain solves problems and getting computers to work in the same way. They’re part of the larger Future Emerging Technology (FET) and Information Society Technology (IST), some of Europe’s heaviest hitters in computation.

FACETS’  neural hardware simulates the processing power of the brain’s neurons  and synapses. The next stage of the chip will act like 200,000 neurons  and 50 million synapses.

FACETS’ neural hardware simulates the processing power of the brain’s neurons and synapses. The next stage of the chip will act like 200,000 neurons and 50 million synapses.

Why do we want computers that think like brains? Electrical signals in computers are faster than neurons, but brains have better strategies at solving big problems. You don’t race your calculator in addition competitions, but you also don’t ask a computer when you should get married. If we want computers that emulate human brain intelligence, we can’t just use brute force (make computers that are faster and more power hungry). We need to make computers that work in a completely different way from the ground up, replicating the amazing power, efficiency, and parallel processing of the brain.

Meet the FACETS

FACETS has to clear three major hurdles in order to reach its goal of making computers think like humans: understand the brain, simulate the brain, and make an artificial brain. To vault these hurdles FACETS relies on different specialists: neurologists, experimental biologists, computer specialists, physicists and electrical engineers.

“The sensory neocortex is considered here as a content-dependent  reconfigurable computing machine.” – Yves Fregnac

“The sensory neocortex is considered here as a content-dependent reconfigurable computing machine.” – Yves Fregnac

One of those experimental biologists, Henry Markram, is studying how the higher-reasoning portion of the brain, the neocortex, divides itself into working groups. He calls these sections of neurons and synapses neocortical microcircuits. They are like tiny sections on an IC chip, each pursuing a different task. Markram and others have included their results in a raw database that anyone can access and use. Find it here. The database is stored in the aptly named Brain Mind Institute in Lausanne, Switzerland. Markram’s work is also part of the Blue Brain Project, an attempt to reverse engineer the brain.

Other biologists, like Yves Fregnac, use their work in FACETS to better understand how neurons and synapses form networks. Yes, even FACETS needs its IT gurus, though the systems these guys study make an office computer cluster look like two tin cans with string. There are millions of neurons and hundreds of millions of synapses in the neocortex, yet FACETS approaches them like any other series of linked computers:

Never say that FACETS doesn’t have a sense of style. Here they show you their simulation processing towers overlapped by images of neurons and synapses they are modeling.

Never say that FACETS doesn’t have a sense of style. Here they show you their simulation processing towers overlapped by images of neurons and synapses they are modeling.

FACETS members Wulfram Gerstner and Alain Destexhe work to simulate the brain in VR. The simulations use massive computer arrays to keep track of millions of virtual neurons and synapses. There’s never a lot of flash and excitement from computer simulations, but the work Gerstner and Destexhe are doing is pulling FACETS closer to understanding how information is formed and moved in the brain. They can watch as neurons fire in their VR world and know they are watching thoughts. Even if they are simulated thoughts.

They Saved Hitler’s Brain! — and put it on a computer chip

Well, FACETS hasn’t gotten to the point of reanimating genocidal maniacs, but they have created a piece of computer hardware that can solve problems in the same style as your neocortex. The so-called “brain on a chip” is really stirring up interest since it made the press rounds in March. Karlheinz Meier, a physicist at Heidelberg, helped create the computer chip that acts like a cluster of neurons and synapses. Using transistors and capacitors, Meier has turned silicon into grey matter.

The FACETS chip has more than 350 designated “neurons” and 100,000 “synapses.” The neurons are made of more than 100 integrated components while the synapses require about a fifth as many. This is the first stage in the project, the next will be the equivalent of more than 200,000 neurons and 50 million synapses. That chip will actually take up an entire silicon wafer. Brains on chips run up to 100,000 times faster than their biological counterparts, and are millions of times faster than simulations.

When are we ever going to use this?

Ask anyone from silicon valley and they’ll tell you, your average silicon wafer is far from defect free. FACETS brain on a chip, then, would seem destined for serious debugging. The defaults, however, are part of the project. Unlike your standard computer, the brain has a large tolerance for mistakes. Kill a single neuron, fry a single synapse, and the brain will keep chugging right along. The same is true for a brain on a chip. If computer chip manufacturers used a brain-like design, they wouldn’t have to scrap parts of every wafer.

And that’s just one benefit of the neural hardware being developed by FACETS. Brains process problems in parallel, they are the computational equivalent of an MMORPG. Millions of different neuron clusters (those wily neocortical microcircuits) are questing after solutions independently. FACETS hopes to bring that kind of computational approach to the mainstream through their work with brain-like hardware.

Over the next year, FACETS original grants will be coming to an end, but their determination goes on. The Marie Curie Institute will help fund PhD programs in several major European universities that focus on brain-like computational paradigms. Each institutional member of FACETS has also pledged to continue its work at least through 2010. Their plan is to develop a network of “brains on chips” and to use this network to help develop a new way of computation. As FACETS develops these increasingly complex circuits that mimic brain-like behavior, we may see the lines between human and artificial intelligence grow thinner and thinner.

Read the full article



En kommentar till “FACETS. Making Computers Work Like Brains”

Kommentera:

Namn *:
E-post *:
Webbplats:
Kommentar *:
*