Chris Eliasmith Discusses SPAUN's Artificial Intelligence

Monday, January 21, 2013

spaun ai engine

Artificial Intelligence
Recently Nikola Danaylov of Singularity 1 on 1 interviewed Chris Eliasmith and discussed a variety of topics including the story behind his desire to create the breakthrough brain simulation, SPAUN.  Eliasmith is working on the Semantic Pointer Architecture as a new model of the brain and is using SPAUN to explore the idea, and will also soon be releasing a book, How To Build A Brain, that will outline the concept.
In November last year, the system SPAUN was announced as major breakthrough in brain simulation. Created by Chris Eliasmith the director of the Centre for Theoretical Neuroscience at University of Waterloo SPAUN,which stands for Semantic Pointer Architecture Unified Network, is a computer model that can recognize numbers, remember them, figure out numeric sequences, and even write them down with a robotic arm.

Recently Nikola Danaylov of Singularity 1 on 1 interviewed Eliasmith and discussed a variety of topics such as: the story behind his desire to create a whole brain simulation and the hardware requirements to run it; whether SPAUN has thoughts and feelings and how would we know if it did; the ethical issues behind creating a brain-in-a-vat artificial intelligence; the relationship between philosophy and engineering; his upcoming book How to Build a Brain; Eliasmith’s thoughts on Deep Blue, Watson, Blue Brain, SyNAPSE and Ray Kurzweil‘s How to Create a Mind: The Secret of Human Thought Revealed; and his take on the technological singularity.

Eliasmith's work on SPAUN is pointing to the need for embodiment to generate an artificial brain.  "Ultimately, I suspect you are going to have a difficult time of actually doing this well if you do not put it into a body.  And so creating a brain-in-a-vat might just really end in the thought experiment that it is.  You might not have a successful  adaptive, effective brain if you do not put it in a body." Eliasmith adds, "So, robotics is definitely something that is important for us in trying to make sure these models are sufficiently robust and sort of...animal like, to explain the kinds of behavior that we are interested in."

When comparing SPAUN to other systems like Watson, Eliasmith suggests that the SPAUN approach is different in that it exhibits a variety of behaviors, and more importantly it cannot switch between these tasks. This essentially means that Watson is not as brain-like as SPAUN. Watson and Deep Blue are highly specialized, and the SPAUN model is not.  As a path to Artificial General Intelligence, Eliasmith's approach is therefore potentially very important.
The basis of SPAUN is Eliasmith's Semantic Pointer Architecture hypothesis. Briefly, the semantic pointer hypothesis states:
Higher-level cognitive functions in biological systems are made possible by semantic pointers. Semantic pointers are neural representations that carry partial semantic content and are composable into the representational structures necessary to support complex cognition.
The term 'semantic pointer' was chosen because the representations in the architecture are like 'pointers' in computer science (insofar as they can be 'dereferenced' to access large amounts of information which they do not directly carry). However, they are 'semantic' (unlike pointers in computer science) because these representations capture relations in a semantic vector space in virtue of their distances to one another, as typically envisaged by connectionists.

SPAUN can download at

According to  Eliasmith “We Have Not Yet Learned What The Brain Has To Teach Us!”

Chris Eliasmith

Eliasmith is also head of the Computational Neuroscience Research Group (CNRG) at the Centre for Theoretical Neuroscience at Waterloo. This group is developing and applying a general framework for modeling the function of complex neural systems (the Neural Engineering Framework or NEF). The NEF is grounded in the principles of signal processing, control theory, statistical inference, and good engineering design. It provides a rational and robust strategy for simulating and evaluating the function of a wide variety of specific biological neural circuits. Members of the group applied the NEF to projects characterizing sensory processing, motor control, and cognitive function.

Some members of the CNRG have recently begun developing applications of related principles to problems in machine intelligence. Specifically, they are constructing novel methods for automatic text understanding that can be used to support classification and clustering.

SOURCE  Singularity Weblog

By 33rd SquareSubscribe to 33rd Square