Scientists at Centre for Theoretical Neuroscience, University of Waterloo, Ontario have successfully created a virtual human brain that can do some of complicated tasks like copying a drawing, image recognition, counting, answering questions etc. The brain requires 24 gigabytes of RAM to work and is powered by a Linux-based supercomputer. Even with this sheer power, the machine takes 2.5 hours of processing for one second of simulated time.
This brain is a virtual model inside a computer, and has around 2,5 million virtual neurons structures in a pattern resembling the overall human brain’s anatomy, including cortical regions, motor control regions, etc. The model is capable of processing visual inputs and store them in a visual memory and can also control a motor to move an arm and draw things. The brain has been named Semantic Pointer Architecture Unified Network, or “Spaun” in short.
The program is more complex than its appears, and scientists have recently published a paper on the Science Magazine to explain its technicalities. But for our readers, we have got some wonderful videos that can clearly explain the working. Like, this is the introductory movie of Spaun.
And here we can see Spaun to reproduce a hand-written digit, exactly in the same way it has been written.
Something more complex, the task now involves Fluid Reasoning, which can give us an idea of its IQ.
The giant complexity of is processing requires a massive super computer to work. The backend of Spaun is Orca and Kraken of the SharcNet High Performance Computing Consortium, a Linux powered supercomputer.
Artificial Intelligence is expanding everyday, and given the power of Linux to handle massive and complex data with ease, and the freedom to customize it to own own's need, it will become the correct choice for AI applications in future.