Printer Friendly

Computer brains with an eye on the future.

With Intel and AMD both announcing 1GHz processors this month we have become accustomed to expecting our computers to get faster and faster.

The disk access speeds, network transmission speeds and raw processing power of much of what this column reports might lead us to believe that it's not long before computers can replace humans.

IBM's efforts with a series of supercomputers bearing the name Blue show computers can just about keep up with the reasoning of a chess grandmaster.

But our computers are extremely primitive when compared to the human brain.

Just consider for a moment the amount of processing the brain has to do instantaneously in order to recognise objects from the impulses sent along the optic nerves. Our brains are even capable of recognising objects when seen from a different angle or when partially obscured.

Now the human brain operates at much fewer operations per second than a microprocessor, but the software can operate very complex pattern recognition algorithms that overwhelm most supercomputers attempting the same task.

In fact, this capability of the human brain to match instantaneously stored patterns against incomplete or new data has led to a complete branch of computer science and its commercial exploitation - it is known as neural networking.

Computer vision systems, which is computers that can replace the human eye, are the subject of much intense research; much of it of military significance and often classified as secret.

Military satellite and other surveillance systems, as well as missile and aircraft guidance systems, need to be able to sift vast quantities of visual information real time and make life-or-death decisions.

Just like humans do, except the West has become increasingly squeamish about having real eyeballs and real brains (particularly highly technically-trained ones) being shot at, so its military planners are seeking technological substitutes.

Occasionally, snippets of information surface from the world of computer vision research that alert you to what is possible. Outside military applications, computer vision has widespread medical application and in biologically hostile industrial environments.

Researchers at Berkeley University in California have developed a hybrid analogue/digital computer called CNN - Cellular Neural Network - which they claim is three times faster than anything else available.

It consists of an array of analogue processing circuits, each of which takes input from an external sensor but also from each of its neighbours, the four on each side and the four at each corner. Signals are amplified or diminished depending on the weight of interconnections between the circuits.

Early tests show CNNs to be up to 1,000 times faster than a Pentium II and faster than conventional digital signal processors. DSPs are special purpose processors designed to handle a narrow range of ultrafast instructions on real time data streams. One drawback is that the analogue devices cannot be scaled down in size at the rate that DSPs can.

More amazingly, a French outfit called Bureau d'Etudes Vision has popped up with a GVPP - general visual perception processor - which it claims mimics the human eye with its millions of rods and cones sensitive to different colours and intensities and neural preprocessing for detecting movement, whereby the brain is not overwhelmed with a pixel by pixel signal but merely with a representation of colour, shape, contour and movement ready for rapid pattern-matching to occur.

BEV claims its GVPP can be manufactured in volume for about $6 and complete vision systems for under $1,000. This compares to the $10,000 currently required for systems with a fraction of the power and functionality.

The GVPP crunches 20 billion instructions per second (compared to the mere millions of instruction per second of Pentium-class processors). Price Waterhouse Coopers will be auctioning the rights to GVPP later this month and estimates a multibillion dollar revenue stream.

Brian Prangle is new product strategy manager at SCC (Specialist Computer Centres).
COPYRIGHT 2000 Birmingham Post & Mail Ltd
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2000 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:Interactive
Author:Pringle, Brian
Publication:The Birmingham Post (England)
Date:Mar 14, 2000
Words:639
Previous Article:Myths of Net commerce.
Next Article:A space for small IT firms to grow.
Topics:

Terms of use | Privacy policy | Copyright © 2020 Farlex, Inc. | Feedback | For webmasters