We’ve reported on a lot of large-scale brain simulations in the past including a partial mouse brain, a rat neocortex and (maybe) a cat brain. None of those simulations actually did anything. Their goal was to simulate the neural network but nothing more. SPAUN is something different. The name stands for Semantic Pointer Architecture: Unified Network. The SPAUN simulation is described in the recently published paper, SPAUN: A Perception-Cognition-Action Model Using Spiking Neurons (PDF format). One of the authors, Chris Eliasmith, has a book coming out soon that details the Semantic Pointer Architecture (SPA) in more detail and describes its basis in the Neural Engineering Framework (NEF). From the paper:
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, 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.
As with any model, it’s not as cool as the real thing. In SPAUN’s case, the model doesn’t learn synaptic connection weights, those were derived by the researchers. The SPAUN simulation has only a single fixed “eye” and a single two-jointed arm. Further, SPAUN can only perform tasks related to series or lists of numbers. Still, SPAUN is an entire working neural system that includes visual perception, cognition, and motor action, which represents a useful advance in the field of brain stimulation. Continued work on this type of model will undoubtedly shed more light on human cognition as well as robotics and AI. And you knew we couldn’t stop with just a description of something this cool, so read on to see some videos of SPAUN actually doing its thing.
SPAUN demonstrates rapid variable creation
SPAUN demonstrated fluid reasoning on an IQ test question
SPAUN demonstrates serial working memory
You can see more videos of SPAUN on the Nengo Build a Brain website.