Skip to comments.Artificial brain 'Spaun' passes IQ tests
Posted on 12/03/2012 9:13:28 AM PST by LibWhacker
Researchers at the University of Waterloo, Canada, have built the worlds most sophisticated artificial brain to date.
Known as Spaun, (short for Semantic Pointer Architecture Unified Network), the model is able to simulate complex brain behaviour and is capable of thinking remembering, seeing and interacting with its environment using a mechanical arm.
Although Spaun currently exists in the simulated world inside a computer, it is one of the most advanced models ever created by scientists to understand how the human brain functions.
Unlike previous brain models, such as the 'million processor computer', which feature large networks of neurons with generally limited functions, Spauns 2.5 million neurons are designed to mimic our own brains structure with a prefrontal cortex, basal ganglia, and thalamus that allow it to think about its environment and respond to the patterns it encounters.
"It has been interesting to see the reactions people have had to Spaun," says Chris Eliasmith, a professor in the department of philosophy at Waterloo University.
Even seasoned academics have not seen brain models that actually perform so many tasks. Models are typically small, and focus on one function."
A collaboration between neuroscientists and software engineers, Spaun receives its visual input via a 784 pixel digital camera this is relayed to its thalamus which performs the same function it would in a human brain: processing the data it receives.
Once processed, the data is stored in Spauns neurons and directed to its basal ganglia which selects the appropriate action it should perform. Using this system, Spaun has been able to perfom well in a variety of IQ-testing tasks, including a test which demonstrates its understanding of numerical concepts.
However, Spauns mental abilities are not something to get too worried about just yet. As Eliasmith explains, theyre still a long way off being able to replicate anything like the sorts of tasks the human brain is able to perform.
"People are enormously more complicated," he says, "[Spaun] is nowhere near as intricate or sophisticated as human brains its miles away."
It is hoped, however that Spaun could provide insights into what happens when parts of the brain become diseased and stop functioning. The researchers have already looked at what happens when you kill some of the synthetic neurons.
"There are not only deep philosophical questions you can approach using this work such as how the mind represents the world but there are also very practical questions you can address about the diseased brain," Eliasmith said.
I believe that critical innovations are going to come from basic research like this. I cant predict what specific industry or company is going to use this work or how but I can list a lot that might."
Videos for Spaun simulations
Work on this is incremental, but it’s progressing. I expect viable android brains within my childrens’ lifetimes.
Always one shortfall....It’s not living!!!!
it’s a simulation, not an actual construct. in a simulation, things occur in a predicted manner, in the real world there is the unexpected.
this is an accomplishment of it;s own, but it still doesn’t exist in the real world.
yeah, it all sounds like so much fun
but just wait until it becomes self-aware
Lest one changes the definition.
No need to go all hyper on fear of a human like brain anytime soon. In order to put it on a chip, you need to have an adequate model of the human brain. Neuroscience is no where near where it needs to be. This is why AI has largely stalled. We were supposed to have androids walking amongst us by now, according to the predictions of the AI community in the 1960’s. They have grossly underestimated what it takes to model the brain.
Once the model is made, creating the hardware is easy. The problem is with the definitions of “human” action and thought that the neuroscientists are using.
Basic philosophical principle: You can’t give what you don’t have. We do not have a good understanding of “understanding”. Most people walking around today have no idea what their purpose in the world is, so I doubt a few scientists could give that ability to a machine.
The only reason that a Skynet may happen is because whoever programs it would have put that seed in there. It would not be able to determine it’s own purpose. Even our children will not see anything more complex that the pint-size companion from the 1984 Buck Rogers TV series.
If you design a machine specifically to pass a test, it will pass a test. As Big Blue with chess. I’ve seen nothing to convince me that AI is still not a chimera.
It is not the “how” that makes the model. You need the “why”, which is philosophy, not science.
Kurzweil’s idea could create the proverbial “tabula raza” brain: capable of processing all the thought that a human could process, but has no inherent idea as to its purpose in the universe except for what the programmer puts into it.
Kurzweil has not solved the perennial “mind-brain” duality question. He thinks that the human brain is just a chemical computer, and if he can duplicate the processes sufficiently it will produce the “mind”, but you cannot produce a “mind” without its purpose and place in the universe being determined.
People do not create a sense of purpose on their own; purpose by its nature is something given to the person. Any purpose you can create for yourself can be just as easily dismissed. Anything so easily dismissed was never really important to begin with. It is because purpose is given from without that it has meaning and importance to us. We do not regard things in our life as important because we made a decision (wife, children, bacon, etc.) but because it was put into us and we recognize something wrong when people do not consider these things important.
So it is with machines. Machines througout history have had their purpose given to them by us, their creators. A machine mind would need a starting point for all its projected sense of morality. Be very wary of whoever programs that first machine “mind”.
How does one ‘pass’ an IQ test?
Beyond the philosophical questions, the human brain seems to have a storage capacity beyond what could simply be measured by a cell-count alone. I think I read somewhere the brain has a trillion cells; I don’t know if that is right, but let’s use that figure as an example. In this digital video age, we know that a single movie can require from megabytes to gigabytes for storage, depending on resolution and compression. How many movies (to say nothing of life events) have you watched that you can replay in your mind? Of course the mind is analog, but how much storage must it require? Is that few pounds of gray matter alone really capable of such storage, or is the mind simply a terminal connected to a larger storage elsewhere?
It’s technically possible to program a computer to play chess flawlessly. It would just take millions of years to make a move.
What is the passing score on an IQ test?
I agree 100% They, computers, do what they are programmed to do and that’s it. No one has a clue how to write a program that will confer intelligence or understanding or self-awareness to a machine. Not even close, imo. The human race is quite safe — for now.