The Most Complex Artificial Brain Yet

By on December 13, 2012 in News 10

Picture: Wikimedia Commons, by Alejandro Zorrilal Cruz (CC)

University of Waterloo’s recent triumph of neuroscience and engineering, Spaun (Semantic Pointer Architecture Unified Network), is the most complex simulated brain yet, consisting of 2.5 million “neurons”. For reference, a non-simulated, non-artificial human brain has one hundred billion neurons, with an average 7,000 synaptic connections to other neurons (100 to 500 trillion synapses in a healthy adult, or 1 quadrillion synapses in a growing toddler).

But by using a small handful of very basic tasks (such as pattern recognition) Spaun can flexibly respond to ‘visual’ input and recreate a sequence of numbers to pass a rudimentary  IQ test. And the research team want to allow Spaun some autonomy through “adaptive plasticity”, to learn and rewire its own neurons without pre-programming.

via  at Extreme Tech:

The New York Times, meanwhile, delves into this idea of ‘deep learning‘, a field of machine learning where ”higher-level concepts are derived from lower-level ones, and the same lower-level concepts can help to define many higher-level concepts,” and work out how many levels of hierarchy best fit the data. Demonstrations around the world are showing the scientific strides in speech, facial and pattern recognition, chemistry, and perhaps even better self-correction, that come from allowing a computer to “brain train.”

But it would be wise to be patient, skeptical, or cautious. The Robot War for the Future notwithstanding, it seems every decade or so that the Singularity is Near, and we should get excited or hunker down for full machine consciousness. Unfortunately, pioneering technology is perpetually just around the corner, and besides, we don’t fully understand our own brains. It was just this year that scientists finished a comprehensive map of the brain. We’re still limited in our understanding of animal intelligence, collective intelligence, what IQ tests are testing for, learning, the origins of our own peculiar brain evolution, what genes code for intelligence, and of course the easy answers to the nature of consciousness itself. New discoveries, interest, and funding, however, are turning out to be very enlightening.

  • BuzzCoastin

    if evolution is correct
    the human brain took billions of years to evolve
    now I read that some advanced chimps are going to build a computer replica of the brain
    how cute

    • Bob

      human beings have hijacked evolution. perhaps natural mindless and purposeless selection has ironically and paradoxically given rise to a species that can give rise to intelligently designed life forms and systems with purpose of form and function.

      • BuzzCoastin

        humans haven’t highjacked evolution
        they don’t even understand how it works yet

        I’m all for exploration & experimentation
        but the rapid implementation of untested technologies
        like gene splicing or electromagnetic fields created by technology or even TV
        have not been deployed with perspicacious intelligence
        but with motivations based on hubris & greed

        from a laboratory perspective
        it’s difficult to distinguish the actions of Capuchin monkeys & humans
        by simply looking at data

        • Bob

          I said “perhaps”. And lets face it human beings are moving beyond natural selection into artificial selection. We are using technology as an evolutionary medium itself. Human beings used to be exclusively manipulated by nature but now we are beginning to manipulate nature. We have manipulated external hard nature the earth, and now it is time to manipulate our soft internal nature; the human body. There will be both positive and negative experiences and occurrences. We are propelled forward whether we want to or not.

          • BuzzCoastin

            > We are using technology as an evolutionary medium itself.

            I think it’s highly possible that technology is using us for it’s evolutionary purposes
            and not the other way around
            wee have become the sexual reproductive organs for machines

            humans have little or no comprehension of the effects of these technologies on humans
            they seems to be deployed with reckless abandon
            this, as the proverbial warnings suggest (e.g. Atlantis)
            is the high road to the collapse of civilization as wee know it
            wheel see

          • Apathesis

            So true about technology being an evolutionary medium.

          • Simiantongue

            There is an underlying assumption there that humans are not natural. Are distinct from or apart from nature. Is that true?

            I’ve always been intrigued by the term artificial. In a basic sense it means “created by humans”, not natural or fabricated and that comes with the connotation “counterfeit”. But I wonder sometimes why something a human does is considered not natural.

          • Bob

            It is merely a distinction of terms between what is created over long geological time (natural) and what humans manufacture in a matter of minutes or hours in a factory (artificial). Human beings and our creations are in no way cut off from the earth. Humans are natural and everything we do and make is natural, because it can occur, and it takes place within Nature which is the entire cosmos.

          • Simiantongue

            I’m not being argumentative, just thoughtful. So please don’t take offense at anything I’m saying.

            I don’t think time is a factor. Geological or otherwise. A birds nest is made in days, it is considered natural. If I were to construct a house, that is not considered natural. However, the materials for the nest and the house come from the same place, the environment. Houses are also made of wood as a nest is. Houses also have sand and rock, even the paints and asphalt are mixtures made from materials we collect much like a bird collects various materials. Surely a house is more complex than a nest but at a basic level the it’s the same premise. The purpose of houses and nests are even much the same. But we make the distinction between the “natural” nest and what is created by humans, the house is “artificial”.

            I’m thinking the distinction between natural and artificial, broadly, isn’t made for any type of utilitarian purpose. The differences between nests and houses are readily apparent, we know that one was constructed by humans and one was not, any descriptor that makes that distinction simply for purposes of utility seems superfluous.

            If humans are in fact part of the environment and are part of the natural world, so are houses, which are made by humans. As natural as any birds nest.

            What I’m thinking is that the distinction is made for other reasons. Having to do with putting humans front and center in the scope of things perhaps. A type of egocentrism. There seems to be another purpose to differentiate what humans do as separate from nature. It seems the term artificial is part of a cognitive bias construct. A purposeful perceptual distortion.

            I can see how that could prove advantageous. It could support ideas like a kind of manifest destiny of man over nature. So rather than being a utilitarian distinction as a descriptor so that we know the difference between bird nests and houses, which is pointless really, it’s more likely that the purpose of terms like “artificial” came about to quell dissonance in how people think of the order of things. To accentuate the things that humans do, aggrandize in thought.

            Humans do have a habit of thinking of ourselves as the center of the universe. It’s not surprising to find that our vocabularies and thoughts reflect that I guess.

  • alizardx

    The IBM project is a *full scale* emulation that runs at 1/1532 full speed based on 96 racks full of state-of-the art customized 17 core processors. I suspect the U of Waterloo project could be implemented in a more powerful way as software in that IBM data center.

    Based on Moore’s Law, I’d say we’re 20 years from hardware that can run human-level sentient AI. The recent google emulation of part of the human visual cortex with the promise of full emulation in 5 years points in the same direction. At that point, software becomes an interesting issue.

    Don’t hold your breath about a Singularity unless you always wanted to look like a blue-skinned Smurf.