Computing Robotics Computational Neuroscience Theoretical Neuroscience

Technology Readiness Level: 8

*As assessed with the Human Brain Project's TRL Guide

Technology Description

SpiNNaker –the acronym of Spiking Neural Network Architecture– is a massively-parallel brain-inspired neuromorphic computer for large-scale real-time brain modelling applications. It has three aims:

To simulate very large brain-like networks, to advance our understanding of how the brain works

As a real-time neural simulator that allows roboticists to design large neural networks, that are both flexible and low power, into mobile robots

To question the restrictions that we impose on our computer architectures, by comparing them to the very different principles evolved by nature in the brain

  • Simulations employing massively-parallel spiking neural networks that mimic the functioning of a brain are run as tools for both computational neuroscientists, to help understand how the brain works, and roboticists, to design large neural networks into flexible, low power robots.
  • More than 1 Million processors in 1200 boards allows large-scale real-time brain modelling simulations without buying time on a supercomputer.

Competitive Advantages

Flexibility

The use of software to model neuron and synapse dynamics allows new theories to be explored rapidly

Scale

With a million processors, each capable of modelling several hundred neurons and several million synapses, real-time models up to full mouse-brain scale are possible

PyNN

A standard spiking neural network description open-access language allows rapid user access with minimal training

R&D

Next SpiNNaker generation will deliver 10 times the computer performance while consuming about the same power as the original chip

Applications and Market Potential

Event-based machine learning for energy-efficient AI, for example in mobile platforms

Neuro-robotic control systems for compliance and user safety

Novel learning algorithms for event-based machine learning

Large-scale brain models, to understand brain function and ultimately, perhaps, to model the effects of drugs

Interesting Facts

  • Around 100 SpiNNaker systems are in use in labs around the world, including US, Japan, Australia and New Zealand
  • The University of Manchester built the world’s first operational stored-program computer, which ran its first program on June 21st 1948
  • Alan Turing wrote his 1950 paper on “Computing Machinery and Intelligence” when at Manchester, introducing the Turing Test for human-like AI – still not passed by any machine!