Through the evolution of time, we can see how a simple electronic calculator has superseded the abacus, which was devised by people in the pre-historic era thousands of years ago. The functionalities of both devices are similar, i.e., a tool for arithmetic calculations. Currently, the use of the electronic calculator has reached far beyond the capabilities and functionalities of its predecessors while preserving its basic set of operations. Complex numerical calculations can be performed using existing scientific calculators that are available in the market. The reason for starting the final chapter with this example is that we can see

that improvements of functionality can be expected from changes that have been made to a device or object that is being used. As has once been said by Marshall McLuhan [127], “A change of medium inevitably leads to a change in the message.” From this perspective, in regards to pattern recognition, we could argue that changes in the way we perform recognition might extend its functions and capabilities. This is relevant to the discussions that have been presented in this book. The way we look at how pattern recognition is performed can be changed if we consider a different approach, i.e., a distributed approach. This book has been dedicated to exploring different ways of implementing

pattern recognition using machine intelligence. Decades of work in improving pattern recognition algorithms have reached their height. It is now the time for us to revisit the fundamentals of biological intelligence, specifically human intelligence. As was mentioned in the preface of this book, the human brain is, in its actual form, a fully distributed network of computational machines, known as neurons. If we are able to map this into existing computational networks, we see that the Internet is actually a physical brain of our entire

computational systems. Just as described in the quote by Dan Millman that opens this chapter. The question that remains to be considered is how we can harvest the functionalities of such large-scale systems. This perhaps can be solved by having computational schemes that work at Internet-scale.