When is Dumb Smart?


IEEE Spectrum Magazine, Nov. 1997


My email this morning is clogged with discussion about whether communication networks should be intelligent or dumb.  Dumb seems to be winning.  Of course most of the participants view themselves as being at the periphery of the network, where they argue the intelligence should be located.  Nobody wants to be the designer or owner of the "dumb" portion.

As engineers we are schooled to provide solutions that are optimized and intelligent.  No test question ever asks you to give a sub-optimum, dumb solution to the problem at hand.  Yet, curiously, in real life being dumb is often the best approach in the long run.  Time has a way of turning today's intelligence into tomorrow's bottleneck.  As the computer scientist Christopher Strachey once said, "It is impossible to foresee the consequences of being clever."

Ah, cleverness!  We seek it, and we engineers are so proud of our clever solutions.  Hank the programmer comes home one evening elated with a small success.  When his wife asks what he did at work, he explains how he was able to save two bytes of memory by truncating the year to only the last two digits.  "Memory is expensive stuff," he tells her.  "Just imagine all the savings if every program uses only half as much memory to express the year."  He is proud of himself then, but as the year 2000 approaches, Hank cannot be found.  No one confesses to being Hank.

I wonder how many of you, like me, have done some really clever programming trick, and have felt the flush of pride that you were so smart to have thought of that.  Then a year or so later, having to modify the program, you can't remember or reconstruct what it was you did.  And the clever thing, whatever it was, seems to be tangling everything up.

I don't know what engineer first conceived of the idea of putting loading coils into access lines for telephony.  This was a triumph of early circuit theory.  The inductance of the coil canceled the capacitance of the line, leading to a flatter transmission passband.  But  now that we need the whole band those loading coils have to be tracked down like escaped criminals at great cost and bother.  The engineer responsible would be tracked down and jailed too, if anyone could find him.

Other people found clever ways to steal bits from the speech signal, or to take advantage of the special characteristics of speech.  Saved money at the time, and a lot of people felt good about their work, got raises and promotions, and ended up running their world -- a world hobbled and unprepared for the coming onslaught of data.

Imagine, though, coming home at night after a mindless day's work of conformity.  "Well," you explain to your spouse, "I designed this really dumb widget.  Basically, I just copied stuff from a standard, and it doesn't do anything special."

Imagine, too, the subsequent performance review with your boss.  "Jones, this design of yours is really dumb, and shows an amazing lack of foresight."

Not exactly sure how to take this, you mumble, "Uh, thank you, sir."

The words themselves are heavy with import.  "Dumb" is bad.  "Intelligent" is good.  On the other hand, "clever" has certain overtones.  We feel pride when we think of ourselves as clever, but aren't so happy when we hear ourselves being described as "a clever person."  "Clever" seems to go too far; as if there is a nuance of scheming.  Perhaps the same thing holds true in technology.  Although we seek to be clever, its actual realization often has consequences akin to that of scheming.

From time to time manufacturers try to make being dumb a virtue.  I remember particularly the fleeting popularity of the "dumb" terminal, which reached its apogee in the time-shared computer era.  For a brief moment, owning a blue terminal, heavily advertised as being as dumb as you could get, was the thing to do.  Now, of course, as a similar idea is being resurrected with the so-called NC (network computer), no one calls it dumb.  That would be a dumb idea.  The implication, and perhaps the truth, is that you have ceded the intelligence and control to others.  That may or may not be a good thing -- it just doesn't sound good.

The issue of the location of intelligence has a lot to do with ownership, control, and innovation.  The real virtue of a dumb network -- of which Internet is the prototypical example -- is that it allows other people to innovate.  This is a great advantage if you belong to the "other people," but may not seem so compelling if you are the network owner yourself.  The network owners argue in the classical tradition of engineers that putting intelligence inside the network, where it can be cost shared, is the most efficient design.

Centralization may indeed be the optimum design.  If this were a college test, it would be the right answer.  But in real life there are a couple of difficulties.  First, there is the assumption that you know in advance what will be needed, or are able to respond quickly enough to changes in technology and applications.  The other flaw is the disproportionate power and flexibility of the periphery.  There are a lot more "other people" than there are network people.  Therefore, more money and more innovation exist at the periphery.  Internet has shown how that power can be unleashed, so long as it isn't inhibited by inherent limitations in network functionality.

The arguments are more complex and less one-sided than I am able to convey here.  For now I am just concerned about the rules of life and engineering, the fragility of cleverness, and the motivations that cause us to be ashamed of dumb solutions.

When I get home tonight, I'll tell my wife that I wrote this really clever column.  But perhaps in a few years I'll have to disown it.  Hopefully some of my dumber columns, and there are many to choose from, will live on.

 

Robert W. Lucky