What's an Engineer Got That a Banker Doesn't?

Many believe that tools are what distinguish man from other beasts. I would argue that equally important is the advantage of a written language and the ability to transmit knowledge to future generations by writing experiences down. Perhaps there have been sheep born with a keen sense of their environment who figured out too late they were destined for the slaughterhouse. Unfortunately for their descendants, they had no way to pass that information on and so the next generation never made a break for it because they just didn't know. When it comes to financial and political progress, it is hard to see how we are any better than the ovines.

If one looks at the progress of man over the centuries, it is clear that technological advancement far outpaces social and financial progress. A case in point is the Ponzi scheme. Charles Ponzi foisted his con on an unwary public (although Wikipedia credits Charles Dickens with inventing the concept in 1857) in 1920 and became a millionaire overnight. Bernie Madoff has pulled off an identical scheme some 90 years later. Investors behaved like sheep when it came to acting on past information. I'll contrast that with Dynamic Random Access Memory (DRAM) evolution, one of the mainstays of modern computer technology.

The DRAM was invented in 1970 by Intel and put into use with the then-nascent microprocessor chips. It was popular because it was the lowest cost storage per bit of any random access technology (and still is). It was cheaper to make because it required the fewest number of transistors per bit, so you could pack more cells in a given amount of silicon. Nevertheless, in the words of the late great Milton Friedman, "There is no such thing as a free lunch." The penalty for this storage capacity efficiency is that DRAM requires its memory cells to be "refreshed" periodically (every couple of milliseconds or so). If not refreshed they would forget whether they contained a one or zero. The need to refresh the DRAM conflicts with the need of the processor to read or write data.  It turns out this was a very tricky design to make work correctly. The major issue is which process gets access when they both try and access the DRAM at the same time. Improper operation invariably causes data corruption and a system crash. Suppose this happens only once in 10 billion operations. That would mean a crash every two hours or so for a microprocessor running at a mere 5 megahertz clock rate. You get the point; this interface must run perfectly to have a useful product.

At some point in time, the major chip vendors decided to fix the problem. They were tired of lousy products being made with those chips inside. They also were tired of the time it took any particular engineering team to finish a design. It was a terrible drag on acceptance and adoption.  Therefore, what they did was design and debug the "definitive" reliable DRAM interface and then integrate it into the microprocessor itself.  In essence, they encapsulated the knowledge in concrete (silicon to be exact) so that no one could redesign it incorrectly. Essentially, they passed the knowledge from one generation of engineer to another in a way impossible to ignore.  In less than 20 years time, DRAM interface went from a major design problem to a non-issue.

Perhaps a closer to home example is the automobile.  In the 1960s (and before), a tune-up consisted of changing the spark plugs, gapping the spark plugs, changing the points, adjusting dwell and timing, and adjusting the carburetor.  All went reasonably well if a good mechanic was involved, but often that wasn't the case. Moreover, performing a tune-up every 12,000 miles was required. Worse still, the car's performance deteriorated from the second the hood closed. By the time mile 11,000 rolled over on the odometer, that tune-up was really needed. Today a tune-up consists of replacing the spark plugs at 100,000 miles. This is done because the manual says so, not because the car is running worse.  In essence, the knowledge of how to do a good tune-up and issuing an alert to have it performed has been passed down to the next generation by virtue of the car design itself.

In comparison to the DRAM of technology, Mr. Madoff, our case in point in the social and financial sector, pulled off a swindle 1,000 times bigger than the original. And this happened despite the fact that millions of people knew what a Ponzi scheme is and despite all the laws put in place to prevent a recurrence.  Clearly something went terribly wrong in the passing down knowledge department between the generations.

We had the crash of 1929 and then the Depression. We had the crash of 2008 followed by who knows what. The reasons for what happened reach far and wide, but "mark-to-market" accounting rules have been postulated by many financial experts as responsible for a fair amount of the blame. Mark-to-market, designed in the 19th century for and by commodity traders, is not a new concept. Put to use in other businesses (notably the banks) in the early 1930s, it was subsequently repealed by President Roosevelt  in the late '30s because economists had figured out it was a culprit in prolonging the Depression. However, the Financial Accounting Standards Board reinstituted the rule in 2007 and the rest is history. Did they learn anything from the previous experience? One would have to say no. Oh, and they just relaxed the rule a few weeks ago, after having wiped out trillions of dollars of investor worth.

What explains the rapid evolution of technology as opposed to finance and social development? Do engineers come from a different gene pool? Do they have better reading skills? Do they eat more veggies or add Wheaties to their diet at an earlier age?  If we do figure out why technology progresses faster can we apply those lessons to financial and social issues?  Or are we stuck with Ponzi schemes forever?

Perhaps first and foremost, technological problems lend themselves to getting a right answer.  And it is easy to figure out when an answer is right because the objective measure is usually straightforward. In the case of DRAM integration, it was how many hours can it run without an error. Three hours was a wrong answer, while 300,000 was the right one. End of story. It is great to be able to say that technique "x" is correct and "y" is a failure.  Meanwhile, we as a species still debate socialism vs. capitalism, mark-to-market rules, executive pay, proper tax rates, etc.

Second, the marketplace is completely unforgiving of bad products. The failure mechanism is a natural occurrence; people just stop buying the product. The manufacturer can either correct the deficiencies or go out of business. That turns out to be a very powerful incentive to get it right.  In the natural course of technological evolution, products go from being innovative to rejected by the marketplace. It is unlikely that buggy whips will reappear in great numbers anytime soon. But why did mark-to-market accounting reappear after its highly unsuccessful debut in the '30s?

The third difference is the way we transfer knowledge from one generation of practitioner to another. There is the written word consisting of books, learned journals, application notes, etc.  This method is common to everyone.  However, technologists have two huge advantages. First, in the case of documents, successful technologies and products have strict revision control, which means that changes as to why and what is done to either enhance the product or fix a bug are recorded. By creating documentation of the changes, the engineer has some confidence that what he is looking at actually represents something that works. Contrast that to the TARP and budget bills recently passed by Congress where many of the members admitted voting for them without ever having read them. There is also the issue of physical reality.  When you design a new system using a released set of parts, you can hook them up incorrectly. But what you cannot do is alter the insides of those parts and negate the work of those who came before you. There is not a whole lot of room for interpretation. Contrast that with the U.S. Constitution, a document subject to judicial interpretation and very much affected by the temperament and mindset of a judge.

In 1929, cars were unreliable and belched all sorts of noxious fumes into the atmosphere. Computers did not exist. Airplanes were slow and unsafe. Movies had just recently gotten sound. TV and FM radio did not exist. The iPod and cell phone were misspellings. DNA was just three letters and genetics non-existent. The stock market crashed and we were on our way to a depression.

In 2009, cars are reliable and produce little noxious gases. There are billions of computers in the world, especially when you count the ones in cars, thermostats, cell phones, etc. Airplanes are fast, safe, and affordable. Home entertainment centers have surround sound and quality so good you think you are in the movie theater. TV is digital and FM is everywhere. Close to 200 million iPods billions of cell phones have been sold. The human genome has been sequenced multiple times, and the stock market crashed.  Many argue that we are headed for a deep recession, if not depression. It seems more than obvious to me that we need to import some of the methods of knowledge transfer from the technologists into the rest of our life structures.

Personally, I think this is a daunting task. First, we need consensus on what works and what does not. Just one week ago, the U.S. was surveyed as to which system is better: socialism or capitalism. Capitalism won by 5%. So if we cannot get a consensus on such a fundamental issue, how do we expect agreement on anything else? Members of Congress enjoy their membership. And because of this, they will do things that enhance their chances for reelection, whether or not they are good for the country as a whole. Apparently, reading bills prior to voting on them is asking too much. We need to revamp our laws and legal system to reduce the number of laws subject to interpretation. Even though Sarbanes-Oxley passed, it has had documented detrimental effects on new company formation. It has also seemingly failed to curtail illegal business acts. More laws will not fix these issues. A rewrite, in combination with careful revision tracking, is in order.

I fully realize that many of the problems we face as a society are far more difficult than the DRAM design problem, but we must start down a different path to address the key issues of today. A well-known definition of insanity is doing the same thing over and over and expecting different results. We really need to change the way we enact policy and the way we write laws and regulations, because we are clearly doing many of the same things over and over.  Otherwise, in the year 2089, the occupants of a mission to Mars, when told that the stock market just crashed, will need to head back to earth immediately to answer a margin call. This lends a whole new meaning to the well-known definition of insanity.