Neuromorphic Supercomputer With 1 Million Cores Mimics the Human Brain

SpiNNaker Credit: University of ManchesterSpiNNaker Credit: University of Manchester

A European neuromorphic supercomputer project that aims to simulate an artificial brain using what's called a Spiking Neural Network Architecture (SpiNNaker) went live on November 2, courtesy of the University of Manchester

Mimicking the Human Brain

The SpiNNaker machine has 1,000,000 neuromorphic processors that are inspired by the human brain neurons. Having 1,000,000 cores was necessary to simulate real-time brain modeling. Each of those processors is capable of completing more than 200,000,000 actions per second and has 100,000,000 transistors.

The neuromorphic supercomputer is supposed to model a biological brain more than any other computer on the planet. A biological neuron is a basic brain cell that communicates by emitting “spikes” of pure electro-chemical energy, according to the announcement. The neuromorphic computer uses large scale systems containing electronic circuits to mimic these same spikes in a machine.

Unlike a regular computer that sends massive amounts of information via a standard network, the SpiNNaker mimics the massively parallel architecture of the brain and sends billions of small amounts of information simultaneously to thousands of different destinations.

Use Cases for the SpiNNaker Machine

The SpiNNaker has 1,000,000 cores, which is an impressive feat on its own; however, that is still far less than what would be required to properly simulate an entire human brain. For comparison, a mouse brain has 100,000,000 neurons, and a human brain has 100,000,000,000 neurons. Therefore, the supercomputer could  need 100,000 times more cores than it currently has to achieve parity with a biological human brain However, it's possible we'll learn that we don’t need 1:1 parity with a human brain to simulate it accurately, or that we could use fewer physical cores with more “virtual cores” (aka threads) that can act as the neurons.

Despite having so much fewer cores compared to even a mouse’s neurons, scientists still believe that the 1,000,000 core supercomputer could prove useful for modeling certain parts of the brain in real-time, as well as for other tasks.

Professor Steve Furber from the Computer Engineering division at the University of Manchester, said:

“Neuroscientists can now use SpiNNaker to help unlock some of the secrets of how the human brain works by running unprecedentedly large scale simulations. It also works as real-time neural simulator that allows roboticists to design large scale neural networks into mobile robots so they can walk, talk and move with flexibility and low power.”

History and Future of the SpiNNaker Project

The SpiNNaker project was 20 years in conception and 12 years in construction at the University of Manchester’s School of Computer Science. They started building the machine in 2006. It initially received £15 million in funding from UK’s Engineering and Physical Sciences Research Council (EPSRC).

In 2013, the project was dubbed the Human Brain Project (HBP), and the European Union and over 100 academic institutions promised to give it €1 billion in funding. In 2015, neuroscientists criticized HBP leadership for wanting to cut cognitive neuroscience from the initiative and sent a letter to the European Commission in protest. The HBP executive team was disbanded and replaced by a 22-member board of directors from multiple institutions the same year.

Earlier this year, Chris Ebell, the new director of the initiative, quit due to disagreements with the project’s lead institution.

Nevertheless, the project has gone live, and its initial creators hope that they’ll eventually be able to upgrade it from 1,000,000 cores to 1,000,000,000 cores in the near future in an effort to get it closer to the number of neurons that a human brain has.