IP: Unsafe at any Clock Speed

[David Farber <dfarber () earthlink net>]

-----Original Message-----
From: "James H. Morris" <james.morris () cmu edu>
Date: Thu, 18 Apr 2002 11:09:59 
To: farber () cis upenn edu
Subject: Unsafe at any Clock Speed

Dave:
The thread about computers compared to cars prompts me to send the following.

Jim Morris


This week my laptop started freezing up, so I took it to tech support and 
had them look at it. They ran some virus checks and tweaked some of my 
setting and sent me on my way. The connection to my cell phone still 
doesn't work. Sometimes when I boot it up it can't find all the RAM I have 
installed. Sigh. I guess it's time to upgrade.

Does this whining remind you of anything other than your own hassles with 
computers? It reminds me of the 1950's when we were in the golden age of 
automobiles, when a General Motors CEO named Charles Wilson declaimed 
"What's good for General Motors is good for the country and vice-versa." 
The gold in question was all going to the manufacturers because we, the 
post-World War II consumers, were so entranced by our new toys that we 
would never think to demand that they perform reliably. The auto companies, 
the oil companies, and the highway builders created a money pump for 
themselves: create big cars that guzzled gas and encouraged consumers to 
see the USA in their Chevrolets. We didn't mind that pieces fell off cars 
and that many of the cool features didn't quite work. I even remember the 
day that a mechanic told me he couldn't get my engine back together. We put 
up with all this because simply having and automobile was a new, empowering 
experience. Who would quibble about the push-button windows not working 
when you could take your girl to the drive-in movie? Savvy buyers learned 
to avoid buying any model in the years immediately following a major 
redesign so as to avoid the many design and manufacturing glitches.
Later, in the 1970's, things changed. Ralph Nader uncovered serious safety 
problems and became a general thorn in the side of auto companies. His book 
about the Chevrolet Corvair, Unsafe at any Speed, had a big impact. A slow 
down in the oil supply caused a crisis, higher gas prices, long lines at 
service stations, and a premium on fuel-efficient cars. Then the Japanese 
automobile companies attacked the US market with more dependable, 
fuel-efficient cars than Detroit could produce. It was very painful for the 
US manufacturers. Eventually, they raised quality and stabilized their 
market share; but the golden age was over.

Nowadays, automobiles are a basic part of our infrastructure. We use them 
intensively and expect them to work reliably and economically. The 
automobile companies are no longer the kings of industry, but they make a 
decent living and can depend upon steady demand from a country that uses 
cars for everything. The election of a president and vice president from 
the oil patch reminds us that the old system is alive and potent.
Speaking of Nader, I wonder if he appreciates the irony of his role in 
installing two oil-patch boys in the White House. He helped to tame the 
automobile triangle and he might be helpful in taming the computer 
industry, but he seems to have engineered a four-year encore for the 
industry of the past.

We are in the waning days of the golden age of computing. Bill Gates 
typified the times when he said recently "The first version of the .NET 
technology will be rolled out next year with a more robust version 
following in 2002."  Industry pundit Guy Kawasaki puts the idea more 
succinctly: "Don't worry, be crappy." In other words, the consumers of 
computing are so besotted that they will buy things that only barely work. 
In such a world the companies should pursue large market share first, use 
clever to techniques to lock in customers, and pursue quality later if 
their share is ever threatened. Once a large herd of users or customers has 
been corralled they are put on the upgrade cycle. The hardware gets faster 
courtesy of Moore's law and the software gets bigger with more features. 
Early adopters a lured by the new features or¾hope springs eternal¾improved 
reliability. The rest of the herd is dragged into the upgrade by a force 
more powerful than the urge to adopt new fashions: they can't read 
documents sent to them by the early adopters. To read the new documents, 
they need new software. But the new software is too bloated to run on the 
old computers, so they pay for a total upgrade. So we have something like 
the auto-oil-highway triangle: the hardware-software-network triangle.

When will it end? The tech stock meltdown signals that it might be 
happening. Behind the collapse in the bubble is an important technical 
fact: chips are too fast and plentiful for the industry's good. The 
exponential improvements brought by Moore's law are now outstripping the 
software bloat that grows only linearly. Most of us have all the processing 
cycles we need and all the features we can imagine. We can download videos 
and play them on laptop computers. The Internet encouraged many people to 
buy their first computer, but there is no compelling reason to buy a new 
one or multiple ones. In other words, the industry is facing a computer 
glut, just like the more familiar commodity gluts. One expects to see 
futures market for chips and bandwidth soon.

How will we overcome this glut? We had one before in the late 1980's. At 
that point IBM, Digital, AT&T and others were facing the onslaught of 
inexpensive PCs. It looked like new products, PC's, displacing old ones, 
mainframes and minicomputers, but underneath the product types the 
fundamental fact was that computing was becoming a commodity.

We recovered from that glut via the Internet in a kind of post-Cold War 
euphoria. In the early 90's we began to hear about the information highway 
and it was clear, in retrospect, that the country was ready for a new 
technological adventure. Many people purchased PCs to get on the net and 
many companies invested in bandwidth and switching computers to support the 
hoped for traffic. It worked too well. We all overbought and are currently 
working through our excess capacity. I don't think this is a purely 
temporary adjustment, however, the gluts caused by drastic improvements in 
chips and bandwidth will continue.

What is the cure? Some PC manufacturers are struggling to find the right 
colors for computers. More far-sighted ones are developing wearable 
computers so that we might be induced to buy new ones to keep up with 
fashion. But the increase in the quantity and speed of the hardware is 
exponential and cannot me matched by simple, linear increases in the 
upgrade cycle. Certain populations in the third world are growing 
exponentially. Maybe if we could make edible computers for them to ingest, 
the problem would be solved.
In the long term, however, the exponential growth must give way to an 
S-curve. Engineers may be disappointed but the businessmen will find it 
easier. The most obvious path to business improvement is the creation of 
new applications, ones that consume a lot of computing power and satisfy a 
real human need.

Is our imagination up to it? Science fiction might be an agent of rescue. 
Fiction writers can dream up fantasies at exponential rates, so let us 
start mining their ideas. How about time-travel and teleportation? They 
seem to be problematical in the physical world, but in the digital world a 
slightly cheesy version is becoming available. My colleague, Takeo Kanade, 
can make 3D digital movies that can be viewed through the virtual reality 
equipment of my other colleague, Randy Pausch. Suppose some unborn person 
would like to be present at their own birth 50 years from now. No problem: 
simply have the mother and attendants agree to be filmed inside Takeo's 
50-camera studio. The entire event can be recorded from 50 different 
perspectives, and then processed into a 3D movie disk. On the person's 50th 
birthday, he or she slaps on a virtual-reality head mount and can pace 
around the birth studio with the father and the doctors. We can add touch 
and smell to the visual experience, so the person will be able to hold 
themselves as an infant, slipping into the virtual body of the doctor. In a 
similar way, a distant grand parent could be present at the birth while it 
was happening. This is a limited version of time-travel and teleportation. 
You can't go forward in time, and you can only go to places where the 
recording devices have been set up. But with an infinite supply of chips 
and bandwidth, maybe the whole world can be blanketed with reality 
recording devices.

If you quibble that this is not "real" teleportation, consider what an 
ancient Greek would say about our version of flying. He would probably 
quibble because we weren't flapping wings and personally soaring in the 
sky. But flying is still useful and even fun for some.
Another repository of human yearnings is religion that promises 
immortality. Jim Gray has suggested how information technology can supply 
that as well: wire up newborns with enough sensing devices and you can 
capture their entire lives for later replay. Using artificial intelligence 
learning techniques one can build a clones of their brains that will think 
as they would think in new situations they never actually experience. Throw 
this program into a suitably designed robot body and you have a reasonable 
facsimile of the person. This product might disappoint the person who 
really wants to be immortal, but their descendants might pay for it. Rather 
than wondering what great-great grandfather Henry would have done they can 
fire him up and ask him. Note that the robot clone would be a better 
reproduction than a biological clone because it will embody the person's 
life experience.

These ideas are fun to think about and may actually be realized someday, 
but I think the best way to improve computer sales in the near future is to 
make things that work. In other words, simply fulfill the implicit promises 
that the computer industry has been making all along about simplifying and 
improving our lives or increasing our profits.

The lesson of the Y2K experience is not that our computer systems are in 
good shape. There were lots of problems over a multi-year period and some 
serious ones as the odometer rolled over. Why did civilization survive? 
Because civilization is not foolish enough to depend upon something that 
barely works. We build our world on layers of technology. We keep 
flashlights and candles in our houses because the electrical system is only 
up 99.99% of the time. City dwellers don't have wells because the water 
system has been debugged after two millenniums. They don't keep horses 
because our transportation system provides many public fallbacks. No 
individuals and few organizations depend upon computers beyond the 99% 
level. Anyone that did must provide their own electrical generator to feed 
their computers.

Why did some think Y2K would end civilization? They were convinced by the 
computer industry which very good at marketing with fear, uncertainty, and 
doubt. It induced a pre-millennium buying frenzy that is giving many of us 
a hangover. My university rushed to install a new accounting system that 
has caused far more misery than a complete absence of computers would have.

Being a two-nines industry is a comfortable place to be most of the time. 
It's like the toy industry. Purchases are optional as far as survival is 
concerned, but children's clamoring for new toys will keep sales moving 
when times are good. Who wants to be a five nines industry, anyway? The 
only one we know is the water business, and I don't see those guys driving 
BMWs. Furthermore, getting there is very perilous. There is a Geoff Moore 
sort of chasm between two-nines and five-nines. Many people will begin to 
depend on us as if we supplied five-nines and they will suffer. How many 
people have on-line brokerage accounts and no other means of stock trading 
today? If that percentage gets too high the effect of a network crashes on 
the stock market could be devastating.



James H. Morris
Dean, School of Computer Science
412 609-5000
http://www.cs.cmu.edu/~jhm



For archives see:
http://www.interesting-people.org/archives/interesting-people/