General

The Marconi Foundation Celebrates Human Ingenuity

Famous engineers and mathematicians gathered to share their interests and concerns as technology continues to develop at a rapid pace.

by Alex Goldman ISP-Planet Managing Editor [November 9, 2005]

On Friday, November 4, the venerable Marconi Foundation held a symposium and award ceremony, presenting a lifetime achievement award to Gordon Moore and a fellowship to Claude Berrou. Robert Lucky, Chairman of the Marconi Foundation, presided with wit and brevity.

Technologies of the future
One speaker suggested that the continuation of Moore's law will force companies to move beyond silicon for chips and aluminum (or copper) for connectors. Possibilities for future connectors include gold as well as carbon nanotubes. Chips may need to move into the third dimension, becoming cubes or globs.

Another speaker suggested that future chips may have more connections and future processors, just like the human brain does.

We were sitting in front of a gaggle of graduate students from the Columbia Integrated Systems Laboratory who were studying integrated chip design. We asked them, during a lull, whether anyone was looking at replacing electrons with photons, electricity with light, and they said no, that there are several good ideas to investigate first, such as those carbon nanotubes, before the industry needs to tackle one that it has no idea how to solve.

Leonard Kleinrock, known to society members as "the wizard of PowerPoint," said that in any network, there is a point at which bandwidth improvements no longer improve delivery speed, as you hit the wall of latency. The exact limit depends upon network congestion and file size.

A 20 ms delay for transmission across the continental U.S. may not seem like much, but at 1 Gbps, you put at least 20 Mb in the pipe before the recipient computer decides whether or not it wants what you're sending.

You cannot increase the speed of light, and latency will gradually become a more significant problem in communications.

He said that information theory has tackled the theoretical question of how many networks provide optimum transmission. If you diving N Mbps into C channels of N/C Mbps, what efficiencies do you realize? The solution is that a single channel is fastest.

A speaker warned that 10 Gbps over copper, while possible, will have power requirements of 20 times the power requirements of 1 Gbps.

We were also told that at these data rates, "better math" can achieve significant efficiencies. It so happens that the other award recipient, Claude Berrou, was proof of this.

Turbo codes
Berrou invented "turbo codes" with the late Alain Glavieux. The pair proved, mathematically, that signal processing is more efficient if it is massively parallel. Traditionally, Berrou said, a signal was decoded in a series of steps, each involving a different processor. Berrou and Glavieux found that if each processor was constantly exchanging data with all of the others, processing became more efficient.

Berrou is humble and much-liked, as a five minute film by Jennie Bourne showed. The film condensed eight hours' worth of material into a quick overview of Berrou's work.

The invention of turbo codes enabled broadcasters to fill a channel to near the theoretical limit proposed by the Shannon-Hartley theorem.

Since neither Glavieux nor Berrou were active in the field of signal processing—their discovery relied on much-ignored work in information theory—their result was initially dismissed and ignored, but eventually the pair won international recognition and their discovery vastly improved communication, especially in the most valuable, low bandwidth communications systems of all: satellites.

The world we live in
Robert Gallagher, information theorist and inventor of the Low Density Parity Check (LDPC) code, raised social issues in the concluding speech of the symposium.

He said that people no longer understand the technology they depend upon in their home. "Maybe I shouldn't say this, but it might be connected to the rise of fundamentalism around the world."

He called for "talented generalists" to explain what's really going on to people who are not receiving technology news.

Perhaps you, reader, can do this, and certainly I and other tech journalists contribute to it.

With this brief overview, we have been forced to omit many of the ideas covered. All we can say is that if you get a chance to hear any of these people speak, you should take it.

—End