E-mail    Print       Blog

The Blacksmith and the Bookkeeper, Part 3

Contents The Evolving Programmer Rules-based and Human-competitive Machine Intelligence Hyper-human Services Denouement

Thanks, thanks to thee, my worthy friend,
For the lesson thou hast taught!
Thus at the flaming forge of life
Our fortunes must be wrought;
Thus on its sounding anvil shaped
Each burning deed and thought.
- From "The Village Blacksmith" by Henry Wadsworth Longfellow

The blacksmith immortalized by Longfellow, once an historically vital, centrally necessary vocation which enabled social and technological innovation over epochs of human history, is now practically extinct.   The similarly vital role of bookkeeper, which shares much with the blacksmith in terms of ancient historical roots and commercial-enabler, still thrives in the Network Age. The third and final article in this series discusses the evolving  role of computer programmer and speculates on the likely fate of the postmodern programmer -- blacksmith or bookkeeper?

The Evolving Programmer

Do you write code?  Not can you write code, but do you?  When was the last time you cracked open one of Knuth's volumes?  Do you even know who Knuth is?

There was a time when programming meant the encoding of finely-honed algorithms, parsimoniously pruned to the fewest possible instructions; algorithms which would then also have utilized the least amount of memory for execution.  In the early days of digital logic and microprocessors, such approaches were necessary in light of  limits -- what today would be considered to be quite draconian limits.  Imagine, for example, the programming skills that were needed to enable the onboard Apollo 11 Guidance Computer which successfully landed the first men on the moon:

The on-board Apollo Guidance Computer (AGC) was about 1 cubic foot with 2K of 16-bit RAM and 36K of hard-wired core-rope memory with copper wires threaded or not threaded through tiny magnetic cores. The 16-bit words were generally 14 bits of data (or two op-codes), 1 sign bit, and 1 parity bit. The cycle time was 11.7 micro-seconds. Programming was done in assembly language and in an interpretive language, in reverse Polish. Scaling was fixed point fractional.  An assembly language ADD took about 23.4 micro-seconds. The operating system featured a multi-programmed, priority/event driven asynchronous executive packed into 2K of memory."  --  Apollo 11: 25 Years Later by Fred H. Martin, Intermetrics, Inc., July 1994

The vast majority of coding today is ignorant of such constraints.  A 2K limit for even the simplest of applications -- even those written in Java, which was ostensibly designed to minimize an executable's footprint -- would today be considered absurd.  And that's just for the minimalist application.  Never mind the JVM, which is sort of required for anything meaningful to occur.  But an entire operating system squeezed into 2K?  It is obvious that the skills required for a successful programmer in 1969 are very different from the skills required for a successful programmer today.

Perhaps the "wire bender" crowd -- those programmers who today work with embedded systems -- still exhibits traits of the first real programmers.  Embedded systems programming today must still sometimes acknowledge the existence of hard limits, albeit subject to resource constraints vastly greater than those which tested our fathers.   Indeed, embedded systems programmers seem to be quite a different species from the enterprise norm.  But there can be no mistaking any sub-species of the postmodern programmer with those pioneers responsible for the onboard Apollo-era systems.  The pomo coder stripe has evolved into something quite different than what it once may have been.  Generally speaking, the programming skills required to put men on the moon in 1969 would be as superfluous today as would Network Age programming sensibilities have been then.  Since innovation alters the coding fitscape, which in turn places pressure on the resident species to adapt, it should be equally clear that further innovation will have similar results insofar as programming orders are concerned ... which is to say that the evolution of programming must therefore continue.

What does it mean to "write code" today?  Using Netscape 7.1 Composer to author this article, I did sometimes click on the Source tab and make adjustments to the HTML code found there.  Am I writing code?  Strictly speaking, does HTML even qualify as "code?"  If so, then what about the encoding of a particular application using a tool like a spreadsheet?  Variables are specified, functions enjoined, and utility served ... do applications created thusly qualify as such?  Are those that produce them programmers?  If so, then don't similar arguments hold true for (for example) the producer of a PowerPoint presentation? ... or a web page using some WYSIWYG tool?  If not, why not? 

The point is that the nature of programming has changed considerably since those early days, which really weren't that long ago.  Given Moore's Law and ever-reducing resource limitations, the need for really tight algorithms fades; thus, programming has changed.  Given the hard costs and time required to develop custom applications versus the increasing supply of off-the-rack, best-practices-laden business components (which I can lease at a fraction of the development cost), programming changes continue.  A dwindling percentage of the shrinking IT workforce could today even qualify as programmers from the Apollo 11 perspective.  The dizzying speed at which computing technology is changing the world is having a similarly profound effect on the very vocation that is central to its nature.  Just as the blacksmith steadily created custom components for machines that would effectively replace him altogether, the postmodern programmer is coding himself into a corner(-case) job market, destined to fight for the table scraps which the machine-intelligence purveyors will not stoop to consume.

Rules-based and Human-competitive Machine Intelligence

In the previous two installments in this series, the history of the blacksmith trade was discussed, and a rationale for its demise proffered.  A highly-automatable Vocational Knowledge Base (see AVKB in the previous article), which the practicing smithy exposed to a machine-hungry world, made the trade low-hanging fruit from a productivity-enhancement perspective.    Indeed, the unwitting blacksmith was handmaiden to his own demise, just as the postmodern programmer is (perhaps ironically) today.  The fact that the nature of programming is changing even as the demand for programming skills has abated, to serve what is otherwise an increasing global demand for IT-related products, should be a glaring indicator and something of a concern to skeptics; even the computer industry is not immure from the ephemeralizing virtuous cycles wrought by IT adoption.  And of course the dark side of dramatically increasing productivity must be higher unemployment, at least in the short term.  Until innovation gives rise to market demands (and skills to meet those demands) that simply didn't exist before, economic displacement is inevitable.  But there's more that Moore's Law has in store, beyond even the dream of effectively limitless memory and compute cycles ...

Artificial Intelligence ... AI ... may be a red herring.  In truth, the case for pomo smithy cum coder need not rely on anything as speculative or scorned.  But it is interesting to momentarily reflect on the thesis behind the much-maligned AI journey and where it has and may yet lead.

In a 1948 paper entitled "Intelligent Machinery," the stepfather of modern computer science, Alan M. Turing, envisioned a machine that could rival human intelligence; virtually all AI efforts for the past 50 years have leaned on Turing's thoughts.  He identified three broad approaches by which human-competitive machines could be achieved:

1. Intellectual (logic-driven) searches:  Turing's own work (The Turing Machine) laid the foundations for a logic-based approach to intelligence
2. Cultural (knowledge-based) searches:  A body of knowledge (or Expert System) approach to intelligence
3. Genetical (evolutionary) searches: Intelligence discovered via a "genetic" approach

Turing's observation that "intellectual activity consists of various kinds of searches" has set the tone and direction for AI research ever since, with debatable success.  While a human-competitive machine still seems a long way off, the commercial uptake of both logic-driven machines as well as expert systems cannot be denied.  The third approach is probably the least understood and the most compelling, if a truly human-competitive AI system is ever to emerge -- beyond Deep Blue , image analysis and some interesting expert systems.

Eighty percent of Google hits for "human-competitve machine intelligence" also include the name John R. Koza, the Stanford University professor who has been a champion of genetic programming since the mid-1980s.  A logical cousin of genetic algorithms (spawned by John H. Holland, the world's first Ph.D. in computer science), genetic programming environments utilize evolutionary patterns such as crossover, selection, replication and mutation, often resulting in surprisingly intelligent solutions.  A teleology is required -- in other words, selection criteria must specify a goal or problem to be solved.  And sometimes the magic just doesn't work.  But with each successive generation -- constrained primarily by the compute cycles and memory available to throw at the problem -- genetic programming is getting better.  According to the most recent publication from Koza and company, the progressively more sophisticated results produced by genetic programming over the 15-year period from 1987 to 2002 include:

• solving toy problems
• producing human-competitive (creative) results (but not previously patented or patentable inventions)
• duplicating 20th-century patented inventions
• duplicating 21st-century patented inventions
• creating patentable new inventions

1. patentable new inventions may very well be the acme of human-competitive activities -- a Turing Test if ever there was one
2. some forms of creativity and innovation are not exclusively human activities
   
3. Moore's Law all but ensures that genetic programming may effortlessly succeed where other approaches to AI have yet to; it is, perhaps, only a matter of more compute cycles
4. no programming -- or programmers -- are required

Okay, yes, some programming will be needed <sigh>.  A few geeky, old-fashioned hackers (in the Eric Raymond sense of the word) would be nice to have around, if for no other reason than to marvel at the results of our Darwinian black box, and provide a few tweaks here and there.  But nothing like the number of workers that may have been needed to (for example) steady the world's Y2K fears.  Those days -- and the demands of those days -- are gone.  Once the genetic machines are encoded, they would ostensibly even have the capacity to heal themselves ...

We no longer harvest the fields with hired hands.  We have our machines.  And our machines were designed by machines, built by machines, and are sold, distributed and serviced by machines.   No blacksmiths are needed.

It is happening steadily -- to some, it may not appear to be happening at all.  I'm quite sure there were still blacksmith's in 1904 who, when confronted with the informed observation that machines would soon reduce their dwindling numbers to near zero, still scoffed.  A ubiquitous abundance of imagination and absence of denial is not a hallmark of our species; sometimes we are like frogs do tend to succumb to heat when boiled in water slowly.  Though we may never notice our daily rate of aging, we nevertheless grow old.  Change, as such, is inevitable.

Hyper-human Services

So whence goeth the pomo coder?  How will the programming trade evolve?  If indeed early 21st century workforce reductions are harbingers of IT things to come, what skills will be in demand?

Futurist Richard W. Samson has proposed a set of attributes that he believes will be characteristic of "in-demand" trades of the near future.  In a recently published article in the Futurist magazine (September-October 2004) entitled "How to Succeed in the Hyper-Human Economy," Samsom posits that a new class of jobs that involve difficult-to-automate "hyper-human" skills (beyond know-how) will be the successors of post Industrial Age know-how based professions.  What  are hyper human skills?  Those (sometimes ineffable) attributes of being "alive" that are currently very difficult to automate.  Things like:

• creativity and imagination
• love and empathy
• subjectivity
• wanting, valuing, intending
• social skills
• hypothesizing

These hyper-human functions, according to Samson, will become increasingly more valuable as know-how based activities become as automated as agriculture and manufacturing is today. 

If Samson is correct, today's coder will evolve into something more like a concierge than an expert system.  The ability to discern a customer's unarticulated problems will become more important than an accurate implementation of a specification -- at least from a valued skills perspective.  Intuition about a client's needs will replace algorithmic agility as a premium ability.  Big picture knowledge will be valued more than narrow-and-deep expertise.  And the ability to write code of any kind will be about as valuable as the ability to code in assembler today.  Yes, there will be a demand for those skills, just as there is a demand for farm laborers today.  But the demand for enterprise level IT coding skills will be nowhere near what it was at its peak -- which,  in retrospect, was probably 1999.  That party, it seems, is over.

The fate of the bookkeeper in a hyper-human economy may ultimately be just as tenuous as that of the un-evolved programmer, although to a great degree, the bookkeeping function already presumes a basis in hyper-human attributes.  My relationship with my accountant is as personal as that of my priest.  His empathy, discretion and creativity are as important to me as is his topical knowledge.  Arguably, he has already evolved well beyond his know-how based ancestor that kept the books for Longfellow's smithy.   Indeed, the difference between the blacksmith and the bookkeeper was never one of know-how; but rather, the extent to which hyper-human attributes embossed their particular vocation.  It was the extent to which hyper-human attributes were enjoined that, in the end, spelled the virtual extinction of the one versus the evolved survival of the other.

Denouement

This series began with the first stanza of a poem which, when published in 1841, romanticized and immortalized the role of a vital trade that we might have completely forgotten otherwise.  So many other vocations have come and gone without such celebration: dowsers, mule skinners, higglers and scribes have all seen their day, made a living when it was possible, and moved on as needed when economic fitscapes demanded such evolution.  The blacksmith, however, provides a lovely metaphor in so many ways:  self-reliant, centrally-vital to the community, commanding iron and flame with sinew and hammer, the smithy iconified so much of what was and is vital to our notions of truth and virtue.  The metaphor is also salient in that the smithy tended the fields of his own demise.  Programmers today are ironically doing the very same thing, by shepherding the ephemeralizing wake of unrelenting IT virtuous cycles into increasingly broader spheres of economic impact.  The pomo coder is a dying breed, at least insofar as the job market is concerned.  Rapidly increasing productivity does indeed have a dark side; economic displacement is inevitable as innovative job creation is not automatic.

In the end, the blacksmith and the bookkeeper are only convenient metaphors, not to be confused with inherently meaningful symbols.  Any number of contrasting metaphors may have served just as well.  The point of this series, however, has been to examine the diminishing role of the computer programmer in the Network Age, which presumably has only just begun.  In an age in which IT continues to relentlessly explore the boundaries of human existence, the shrinkage of the IT workforce  must be viewed with at least a modicum of interest if not amusement and perhaps a bit of concern.  Hopefully this series has been of some value in that regard.

Max Goff is the Chief Marketing Officer for Digital Reasoning Systems, Inc., headquartered in Nashville, Tennessee.

Related Articles

The Blacksmith and the Bookkeeper, Part 2
This series of articles explores the role of the blacksmith and the bookkeeper in 19th century economies, explains the extinction of the one and the growth of the other, and compares the postmodern role of programmer to both, culminating in forecasts for the likely evolution of software programming as a viable future profession.

The Blacksmith and the Bookkeeper, Part 1
This series of articles explores the role of the blacksmith and the bookkeeper in 19th century economies, explains the extinction of the one and the growth of the other, and compares the postmodern role of programmer to both, culminating in forecasts for the likely evolution of software programming as a viable future profession.

View all java.net Articles.