...the Holy Grail of computer science. From the days of antiquity, people have striven to give their mechanical creations intelligence. Modern work began in 1956 at "The Dartmouth Summer Research Conference on Artificial Intelligence" with the grandiose prediction that,
then the bar was set with a prediction that in less than 20 years they would have,
And here we are, 65+ years later, and the problem has proven to be nearly intractable.
There is no "Artificial General Intelligence" on the market, nobody would seriously consider 'Deep Blue' or 'Watson' to be intelligent: capable of mimicking aspects of cognitive behavior certainly but not intelligent. We have to ask ourselves why. Why have the best minds in the field failed to produce AGI with over 60 years of research?
Perhaps there is, at the root of the question, an assumption so basic that it is nearly invisible? From that seminal conference there has been one assumption prevalent with nearly all AI researchers - that intelligence is computation. If intelligence is computation, then AGI is simply the search for the appropriate algorithms assembled in an appropriate manner and voila! you have an intelligent machine.
So why hasn't it worked? FnnTEK believes that the assumption that intelligence is computation is fundamentally incorrect.
FnnTEK is operating under a different assumption - INTELLIGENCE IS NOT COMPUTATION, IT IS AN EMERGENT PROPERTY - our basic assumption is that General Intelligence (artificial or biological) is a strongly emergent function of the non-linear interaction of a sufficiently complex network of neurons placed into a sufficiently complex environment.
Invocation of 'strong emergence' is anathema to most mathematicians; it feels too much like magic. However, we have been convinced by our early results that our assertion is correct. Our initial system, a binocular vision platform, with a very simple musculature (independent pan / tilt of two cameras) and a small network (~105 neurons) exhibited startling emergent behavior. Initially the system moved the two cameras randomly, which was expected. Within hours however the cameras began to move in concert. A few hours later the cameras would track a moving object and attempt to keep it within the center of the visual field.
None of this behavior was programmed - it emerged from the network without any algorithmic development. This unexpected result has spurred us on to develop what we call a Fractal Artificial Neural Network or FrANN™ and move from a proof of concept laboratory system into several commercial quality Intelligent Products.
Showing posts 1 - 4 of 4. View more »