IBM simulates cat, maps human brain to improve chip tech

Scientists at IBM said they'd made progress towards creating a computer system that will simulate and emulate the human brain for sensation, perception, action, interaction and cognition.

They also claimed that the system rivals the brain's low power and energy consumption.

Scientists at IBM Almaden, in tandem with scients at the Lawrence Berkeley National Lab, said they'd performed the first near real time cortical simulation of the brain exceeding the scale of a cat cortex and with one billon spiking neuros and 10 trillion individual learning synapses.

And if that wasn't enough, IBM has created an algorithm that uses the Blue Gene supercomputing architecture to measure and map connections between the cortical and sub-cortical places in the brain using magnetic resonance diffusion weighted imaging.

Not only will this help to explain the computational dynamics of the human brain, it will also advance a bid to create a small, low power synaptronic chip using nanotech, phase change memory and magnetic tunnel junctions.

Not only will this help to explain the computational dynamics of the human brain, it will also advance a bid to create a small, low power synaptronic chip using nanotech, phase change memory and magnetic tunnel junctions.

The simulation of the cat cortex used a simulator on the Lawrence Livermore Dawn Blue Gene/P supercomputer - it has 147,456 CPUs and 144 terabytes of memory.


IBM is involved in the DARPA SyNAPSE (Systems of Neuromorphic Adaptive Plastic Scalable Electronics) initiative - that intends to build a prototype chip and eventually build low power cognitive computers approaching mammalian scale intelligence.

No comments:

Post a Comment