UNDATED – Chris Eliasmith has spent years contemplating how to build a brain.
He is about to publish a book with instructions, which describes the grey matter’s architecture and how the different components interact.
“Then I thought the only way people are going to believe me is if I demonstrate it,” says the University of Waterloo neuroscientist.
So Eliasmith’s team built Spaun, which was billed Thursday as “the world’s largest simulation of a functioning brain.”
Spaun can recognize numbers, remember lists and write them down. It even passes some basic aspects of an IQ test, the team reports in the journal Science.
Several labs are working on large models of the brain– including the multi-million-dollar Blue Brain Project in Europe – but these can’t see, remember or control limbs, says Eliasmith.
“Right now very large-scale models of the brain don’t do anything,” he said in an interview.
His Waterloo team took a different approach, using computers to simulate what goes on inside the brain, similar to the way aircraft simulators mimic flight.
The clever creation is the first to bridge what Eliasmith calls the “brain-behaviour gap.”
Spaun, which stands for Semantic Pointer Architecture Unified Network, has 2.5 million simulated neurons organized into subsystems to resemble the prefrontal cortex, basil ganglia, thalamus and other cognitive machinery in the brain. It also has a simulated eye that can see, and an arm that draws.
The simplified model of the brain, which took a year to build, captures many aspects of neuroanatomy, neurophysiology and psychological behaviour, says Eliasmith, director of Waterloo’s Centre for Theoretical Neuroscience.
He says Spaun simulates the biological function of real neurons, including the voltages generated in the cells, and the signals zipping around the brain.
“It’s all in a machine, but we’re actually simulating all those voltages and currents down to the level of things you can measure in real cells,” says Eliasmith, noting there are no connections in Spaun that aren’t seen in the brain.
His team reports that the virtual brain can perform eight tasks that involve recognizing, remembering and writing down numbers.
They say Spaun can shift from task to task, “just like the human brain,” recognizing an object one moment and memorizing a list of numbers the next.
And like humans, Spaun is better at remembering numbers at the beginning and end of the list than the ones in the middle.
Spaun’s cognition and behaviour is very basic, but it can learn patterns it has never seen before and use that knowledge to figure out the best answer to a question. “So it does learn,” says Eliasmith.
But it is not – at least not yet – a match for the real thing.
“Spaun is not as adaptive as a real brain, as the model is unable to learn completely new tasks,” the team reports in Science. “In addition, both attention and eye position of the model is fixed, making Spaun unable to control its own input.”
Observers say the Waterloo brain captures key aspects of perception, cognition and behaviour. It sets a “new benchmark” for large-scale simulation of the brain, Christian Machens, of the Champalimaud Centre for the Unknown, a research institute in Portugal, says in a review of the Waterloo work also published in Science. Machens was not involved in the research.
Eliasmith is now working with groups in the US and Britain to try speed up Spaun and expand its tasks and behaviors.
He says such brain simulations might one day be used to better understand and model neurological disorders and diseases and to improve “machine intelligence.”
Eliasmith notes that humans have about 100 billion neurons in their brains, far more than other animals and artificial brains taking in shape in the lab.
“I think what’s special about humans is the number and connectivity of their neurons,” he says, adding that it appears the more neurons available, the more sophisticated the brain structure. “It comes down to the number of resources you have for processing information.”
Today’s “smart” machines can play chess, backgammon and act as personal assistants, like Siri on Apple’s iPhone, but Eliasmith says the processes they use have little in common with the brain.
He says it hard to predict the future, but he expects to see an explosion in artificial intelligence and more “human-like” machines.
“A robot that is able to navigate through a city and deliver a package from one place to another,” he says. “I think that kind of thing will be within reach in the next 10 years.”
Eliasmith’s book, entitled How To Build A Brain, is due out in February.
COPYRIGHT – POSTMEDIA NEWS