Scientists are working on a circuit that mimics the human brain in an effort to dramatically boost artificial intelligence and machine learning.
Researchers at the University of California at Santa Barbara said this week they have created a circuit running about 100 artificial synapses that was able to perform a simple image classification.
“Classical computers will always find an ineluctable limit to efficient brain-like computation in their very architecture,” UC Santa Barbara research associate Mirko Prezioso said in a statement. “This…technology relies on a completely different way inspired by [a] biological brain to carry on computation.”
The university is hoping the circuit could one day be expanded to approach the scale of the human brain, which has 10 15 -- or one quadrillion -– synaptic connections.
Researchers are focusing on building a computer based on the human brain because, regardless of a human’s ability to forget things or other mental errors and the increasing power of computers, the biological brain remains engineers’ model for computational power and efficiency.
“This is going to help them create computers that can learn,” said Zeus Kerravala, an analyst with ZK Research. “You want machines that can take in data and use that to improve the ability to perform tasks – to learn how to do these tasks better. I think these circuits would let computers do this at a much greater scale…. Computers would start acting more human.”
Dan Olds, an analyst with The Gabriel Consulting Group, noted that the UC Santa Barbara work could be the best way to improve computer vision and vision understanding.
“Object recognition is maybe the most promising aspect of machine learning,” said Olds. “When a machine can truly see and recognize objects, it's a short step to allowing them to make decisions and take actions based on what they see. A machine with advanced vision and evaluation capabilities could do most anything a human can do -- from cooking a restaurant-quality meal to picking up garbage alongside the road.”
He added that a garbage company could save a lot of money by having a robotic garbage collector that could pick up the garbage, recognizing the difference between a trash can and anything else it might find on the side of the street.
UC Santa Barbara isn’t the only organization working on neural-based circuits and computers.
A year ago, researchers at Sandia National Laboratories announced that they have been working on creating a computer system that works more like a brain than a conventional computer. The work on neuro-inspired computing at Sandia is being done as part of a project focused on future computing systems.
Last summer, IBM noted that it had hit a milestone in its work on building a brain-inspired computer system. Code-named TrueNorth, the neural-based procecessor has 5.4 billion transistors on a chip network of 4,096 neurosynaptic cores; it produces what would be the equivalent of 256 million synapses.
Olds said this kind of work -- the UC Santa Barbara work specifically -- could lead to interesting new types of computing systems.
“I think the UC Santa Barbara breakthrough holds great promise and could be the best way to get to truly visionary machines,” he added. “I'm intrigued with where this technology is going.”