EXPLORE - EXCEL - ENJOY

BODY - MIND - SPIRIT

Please note that your COMMENTS are moderated.

On small MOBILE devices (e.g. smartphones), the MOBI version of the BLOG
will be automatically displayed to facilitate the reading.

IBM effort to build a BRAIN


from Kononia House
International Business Machines (aka IBM) has set out to tackle an impossible task – creating a super computer that can act like the human brain. The company has succeeded in producing a multitude of artificial neurons and synapses, and the researchers claim to have conquered more processes than are found in a cat's brain. The 4.5 percent of the human brain they have simulated does do amazing things for a computer, but of course there is no way that it can truly compare to the astoundingly efficient, powerful and ever-changing folds of gray matter inside each of our heads. 
The human cerebral cortex consists of about 20 billion neurons in an über complex network connected by about 200 trillion synapses. IBM has succeeded in simulating a fraction of that processing power in a supercomputer that has 147,456 parallel processors, each with about 1GB of working memory. It's 1.6 billion computer neurons and 8.87 trillion synapses work together to simulate the thalamocortical loops found between the thalamus and cerebral cortex in the human brain. The simulations exceed the scale of processes found in the cortex of the house cat brain, and about 4.5 percent of those in the human brain. IBM hopes to add the 732,544 more processes to make up the power of one entire human brain by 2019. 
The computer has a great deal of processing power, but what exactly can these simulations accomplish? 
Specifically, the computational neuroscientists at IBM are working in reproducing the thalamocortical loops that are proportionally so much larger in humans than in other mammals and might be a key to why humans (generally speaking) have higher reasoning abilities than, say, dogs. 
The thalamus acts like a sensory relay station sending signals about spatial sense and motor signals to the cerebral cortex. It also regulates consciousness, sleep, and alertness. The cerebral cortex is the "gray matter" of our brains. It is vital for our ability to perceive the world - sight, sound, taste, smell, touch - for memory, for consciousness and thought, and for language. Descartes was able to say, "I think, therefore I am," because he had a cerebral cortex. 
The IBM researchers have therefore attempted to reproduce sight, memory, and computational ability through their supercomputer, given the rather cold name "C2."  C2 has no eyes, but it does have a "brain-cam", the data from which gets converted into an MPEG movie that can be replayed. It has artificial neurons that fire across synapses and C2's makers claim it has simulated the brains of a mouse, a rat, and a cat. 
The IBM researchers report that the simulations "incorporate phenomenological spiking neurons, individual learning synapses, axonal delays, and dynamic synaptic channels, exceed the scale of the cat cortex, marking the dawn of a new era in the scale of cortical simulations." 
Biological v. Computer Brains: 
As far as computer science goes, this stuff is fantastic. The processing power and memory required to simulate even a mouse brain goes far beyond what can be accomplished with the common laptop. Consider the computational power in 147,456 CPUs and 144 terabytes (144,000 gigabytes) of memory. That's what these guys are working with. 
Biologists flippantly say that the brain is the result of adaptive evolution over millions of years. It works because the creatures whose brains didn't work died. Yet, brilliant scientists modeling their supercomputer on existing biology have struggled to produce anything close to the full power of a living brain. The researchers are pleased to have surpassed the computational power of a cat. Yet, the computer still has no hope of pouncing on rodents in the grass, of computing the exact distance to leap and sending signals to a system of muscles within a fraction of a second. The computer cannot use the vast acres of data taken in by one sweep of a cat's head or prick of its ears or whiff of the air in the exceptionally efficient and effective way that a cat does to know that a vole is hiding in the grass two feet away. The computer doesn't enjoy eating that vole. It has no emotion. It cannot care, and it cannot add more neurons to itself to simulate caring. In fact, it is questionable whether the IBM researchers, with all their biological brain power can figure out how to simulate "caring" in a computer at all. Ever.

These great minds have worked day and night to attempt to recreate merely the processes of the cat and mouse and rat brains with little true reflection of those brains' abilities, and they have given themselves eight more years to get close to simulating a full human brain. Even then, we suspect it will fall dismally short of the reality. 
It is easy to say that the brains of "higher" organisms evolved over millions of years. However, biologists are hard pressed to demonstrate that any such thing is even possible. Neuroscientists don't fully understand how the brain works, let alone how it came to be.  Unfeeling, unthinking, computers may be able to solve complex mathematical problems or pull up items from their memories faster than we can, but they use far more power to do so, and they miss the nuances.  As Mark Fischetti writes in Scientific American, "The incredibly efficient brain consumes less juice than a dim lightbulb and fits nicely inside our head. Biology does a lot with a little: the human genome, which grows our body and directs us through years of complex life, requires less data than a laptop operating system. Even a cat's brain smokes the newest iPad - 1,000 times more data storage and a million times quicker to act on it."
Related Links:

No comments :