Researchers across the world continue to make advances towards using technology to mimic the human brain, and scientists at the University of Waterloo have created their own functioning model that can even pass the kind of basic questions that appear on a standard IQ test. Known as Spaun — Semantic Pointer Architecture Unified Network — the artificial brain uses a supercomputer to model 2.5 million neurons (by way of comparison, the human brain is thought to have up to 100 billion neurons). Spaun features an electronic "eye" that lets it receive input in the form of 28 x 28-pixel images, and uses a robotic arm to scrawl its answers on paper.
At the moment Spaun can perform eight different pre-defined taks, including counting number progressions, reproducing the way in which a number it sees is styled, and recognizing number patterns — the kind of question one would find on an IQ test. Different areas of the brain are modeled to perform the various functions necessary, with the virtual thalamus performing visual processing, for example. This organic approach leads to a system that doesn't necessarily perform flawlessly all the time — in one video, the researchers show how Spaun has problems storing large strings of numbers in its short-term memory, just like humans do — but they say that those failings actually make it easier to determine whether Spaun is accurately modeling the human brain.
The researchers have published their latest findings about Spaun in a recent article in Science. While it's clear that models such as this are still a long way from the kind of artificial intelligence we dream of in science fiction and fantasy, the combination of powerful computing and neuroscience appear to be bringing us towards that threshold at an increasingly accelerated clip.