Giant artificial brain computer unboxed at Sandia labs

Apr. 29—Computer development has taken a wide path away from the silicon-based hardware we've grown accustomed to.

Research has been conducted into various other ways of building even more efficient systems. There are wetware computers, fungal computers, quantum computers, supercomputers, chemical computers and DNA computers.

Now, Sandia National Laboratories has unboxed a newly delivered brain-based neuromorphic computer, which uses artificial neurons for computations — 1.15 billion artificial neurons. Neuromorphic computing is inspired by the structure of the human brain and uses artificial neurons to mimic brain function.

That staggering number of artificial neurons signals the computing power of the machine, which the labs received from Intel Corp. The new system will provide Sandia and the National Nuclear Security Administration the tools they need to solve problems involving large-scale physics, chemistry and the environment, Sandia lead researcher Craig Vineyard said.

"We believe this new level of experimentation — the start, we hope, of large-scale neuromorphic computing — will help create a brain-based system with unrivaled ability to process, respond to and learn from real-life data," Vineyard said in a news release.

It sounds a little like another leap for artificial intelligence, but it's not quite AI yet. What it is, with its bionic brain, is a system well-suited for AI software.

Neuromorphic computers aim to replicate the brain's neurons and their function. But in today's computer marketplace, silicon still dominates — so neuromorphic software needs its own hardware to produce optimum results.

Vineyard said more is being done in-house to fully take advantage of its power.

"Since a system of this scale hasn't existed before, we've been developing algorithms to efficiently use it," Vineyard said. "Brain-based computing functioning through conventional serial hardware hampers the efficiency of neuromorphic programs."

How it works

Neuropathic computing saves energy and computing time by electrically pulsing only when a synapse is stimulated enough. It mimics the brain in this way, using subgroups of active neurons rather than remote storage involving every possible unit, according to Sandia.

Sandia researcher Brad Aimone said, "One of the main differences between brain-like computing and regular computers we use today — in both our brains and in neuromorphic computing — is that the computation is spread over many neurons in parallel, rather than long processes in series that are an inescapable part of conventional computing."

The more neurons in the neuromorphic system, the more complex calculations the computer can perform.

"We see this in real brains. Even the smallest mammal brains have tens of millions of neurons; our brains have around 80 billion. We see it in today's AI algorithms. Bigger is far better," Aimone said.

A billion-neuron system will create the opportunity to innovate new AI algorithms that may be more efficient and smarter than existing algorithms, Aimone said.

'Multiple computer efforts'

As innovative as the computer system is, it's just the tip of the iceberg at the lab.

Sandia has been at the forefront of multiple computing efforts, Sandia spokesman Neal Singer said. Researchers won an an award for creating a program for the fastest supercomputer in the world, which improved the ability to predict climate change over the decades. Sandia researchers also have produced components for quantum computing.

Advancing computing is part of the lab's broader interest, Singer said.

"There's, I suppose, a triple basis underlying the lab's computer efforts," Singer said in an email. "A pure love of science and technology, an effort to create the latest technology soonest to improve the defense of the United States and an effort to commercialize rather than leave cool advances untouched on the shelf, since it was determined decades ago that a healthy economy contributes to the defense of the U.S."

And the possibilities with a neural computer are, well, mind-bending. Vineyard sees smarter soldier gear, better analysis of intelligence operations, better border security and quicker responses to earthquakes. Medically, he sees more rapid medical diagnosis and less expensive drug discovery.

"We might soon see self-driving cars with neuromorphic technology, lane detection, cell phones with voice recognition, smarter watches and refrigerators, more detailed home security systems," he said. "Did the cat run by or is someone in your house?"