Spaun, a working brain made of spiking neurons
The Computational Neuroscience Research Group at the Waterloo Centre for Theoretical Neuroscience has developed a large-scale model of a functioning mammalian brain using spiking neurons. While other large-scale brain simulations have been done by other groups, this is the first simulation to include all the necessary components to actually perform a real task, from perception all the way to motor control.
A recent paper about the project is entitled Spaun: A Perception-Cognition-Action Model Using Spiking Neurons (available online). The paper, by Terry Stewart, Xuan Choo, and Chris Eliasmith, all at the University of Waterloo, presents an overview of the neural architecture, how it functions, and what it can do. Here’s the paper abstract:
We present a large-scale cognitive neural model called Spaun (Semantic Pointer Architecture: Unified Network), and show simulation results on 6 tasks (digit recognition, tracing from memory, serial working memory, question answering, also providing a reasonably limited semantics that the model addition by counting, and symbolic pattern completion). The model consists of 2.3 million spiking neurons whose neural properties, organization, and connectivity match that of the mammalian brain. Input consists of images of handwritten and typed numbers and symbols, and output is the motion of a 2 degree-of-freedom arm that writes the model’s responses. Tasks can be presented in any order, with no “rewiring” of the brain for each task. Instead, the model is capable of internal cognitive control (via the basal ganglia), selectively routing information throughout the brain and recruiting different cortical components as needed for each task.
This is a substantial step forward in both brain simulation and robotics. Neural models before have done this or that part of what a brain does, but never before have all the pieces been put together into a complete model. Now that we essentially have a working brain in software, we can expect rapid improvements, both in our understanding of real brains, and in practical applications for robotics.
A paper about the work has been accepted for publication in Science in the next few months, and a book called “How to build a brain” will provide even more detail. The research group makes available a free software package called the Nengo Neural Simulator, which implements the Neural Engineering Framework upon which Spaun is built.