World's most complex artificial brain ever passes IQ tests

 

Even though the IQ tests are the most basic ones, the news that an artificial brain can tackle them might give some people pause. Does this signal the rise of the machines? The neuroscientists working on the large-scale model simulation of the brain say this version is still pretty simple compared to the real deal, but it is valuable in helping us see and understand how our brains developed over time.
The artificial brain, built by neuroscientists and software engineers at the University of Waterloo in Canada, runs on a supercomputer, uses a 784 pixel digital eye for visual input and has a robot arm to draw responses. It even has a name — Spaun (Semantic Pointer Architecture Unified Network.
So far this sounds like your average robot scientists have been playing with. What makes this artificial brain different is that it's not constructed like a robot, but rather mimics the structure of a real brain. It has 2.5 million simulated neurons that were constructed to pattern the brain's cranial subsystems. This means the artificial brain has a prefrontal cortex, basal ganglia, and thalamus all wired together to work like the wiring of real thing.
The various sections of Spaun are designed the way scientists felt the real counterparts would work. The visual input from the eye is processed by the thalamus, data is stored in the neurons and the basal ganglia direct the appropriate portion of the brain to complete a task. The team designed the brain to perform its computation in the most physiologically accurate way — even simulating voltage spikes and neurotransmitters.
A basic example of the task or computation required of it is when nscientists flashing various numbers and letters at it. These are read by the Spaun's eye and then reads into memory. Another letter or symbol provides the instruction for the basic input (or memory) and then the robotic arm moves as directed.
Speaking of tasks, aside from controlling robot arms and passing IQ tests, just what can this artificial brain actually do? Is it going to take my job? Not yet; so far it only performs eight different tasks that range from copy drawing, counting, question answering and fluid reasoning.
Chris Eliasmith, who leads the research team at the University of Waterloo in Canada, explained to PopSci:

"These artificial brains don't actually do anything. They don't see, they don't remember, they don't recognize objects," he said. "They sit there and generate complex voltage patterns, but those complex voltage patterns aren't tied to behavior."

What has the scientific team learned? Well, first they know that Spaun struggles to store more than a few numbers in its short-term memory. Bottom line, this artificial brain with 2.5 million neurons is actually fairly simple — but what is interesting is by getting it to process the simple tasks it does start to build more complex behavior by weaving the simpler tasks together. This is the behavior the scientists believe show how our brains may have developed over time.
What does the future hold for this giant artificial brain? The research team isn't finished with it just yet. They want to try to create "adaptive plasticity" in Spaun. This means developing the ability for the currently hardwired artificial brain to rewire its neurons and learn tasks by doing rather than being instructed. And in rewiring neurons, it implies the ability for the brain to "heal" itself — an area of significant study for those with brain injuries.
Eliasmith is working on a program where Spaun wouldn't be given specific instructions, just positive or negative feedback — in theory this would lead to the neurons adapting to finding strategies to complete its own tasks.
The research is interesting on its own, but it does have a larger purpose. Eliasmith tells PopSci: "It lets us understand how the brain, the biological substrate, and behavior relate. That's important for all sorts of health applications."
One example of this has Eliasmith working in reverse. In working with the artificial brain he also "killed" some of the synthetic neurons to see what happened. He was able tow watch how performance degraded — all of which could help scientists understand aging and disorders that attack the brain cells.
In testing he has "killed" synthetic neurons and watched performance degrade, which could provide an interesting insight into natural aging and degenerative disorders.
Interested in the development of the artificial brain? Well, there are a few ways to learn more. ExtremeTech has some additional videos that show how Spaun works. Or, if you are feeling very DIY, you can download the Spaun neural model for yourself (it's built on Nengo — a graphical open source software).
You just never know what you are going to find when you start tinkering with the brain — artificial or not.