I'd like to thank both Saul Silverman and Jay Hanson for much more
moderate replies after my recent comments. In general I'm enjoying
the discussions on this list and I'm glad I signed up for it, though I
naively expected a bit more discussion of the nominal topic, the
future of work.
I suppose one's view of the future of work does depend on one's views
on economics and the overall future of the human species, so I can see
a connection. And I suppose the earlier discussion amongst Eva Durant
and various others about who-did-what-to-whom in Soviet Russia is
ultimately about the very different approach to work under the Soviet
system and what that might suggest about work in the future.
But there does seem to be a great deal of blaming going on in both of
these discussions, and I don't find that very productive.
Having just said that, I'd like to make a couple of mildly critical
remarks about economics, and about Soviet economics in particular.
A remark I often quote is by J.A.Campbell, writing about what he
claims is "the central problem in computer science: avoiding or
minimizing the effects of the combinatorial explosion of possibilities
in a search space". I believe that this claim is too modest, the
combinatorial explosion is not just the central problem in computer
science, but it society as a whole.
In the early days of the Soviet Union there was an attempt to match
people to jobs (or tasks) through some central bureaucracy. Of course
bureaucracies don't work very well, but even if they did work,
perfectly, they could not have accomplished that task because of the
combinatorial explosion of possibilities.
In graph theory and computer science the problem of matching workers
to jobs (or any equivalent bipartite matching problem) is called the
assignment problem.
Good modern algorithms for solving the assignment problem are roughly
O(3), which means that they scale up as to the cube of the number of
nodes. Using my aging 120 MHz Pentium it takes about half an hour to
solve an assignment problem with a few thousand nodes. To solve a
problem with a few million nodes would not take 1000 times as long,
but the cube of that, one billion times as long. So there is probably
not enough computing power in the world today to solve the assignment
problem the Soviet bureaucracy set themselves.
OK, this is an oversimplification. But the basic point should be
clear. The organization of society is the kind of combinatorial
optimization problem that is hard to solve. Actually as combinatorial
problems go, it is one of the easy ones, most are not just hard but
virtually impossible. But somehow most economists don't address the
combinatorial explosion. A flaw in the economics curriculum, I suppose.
Unemployment is a good example. One constantly hears governments
talking about job creation, as if there just aren't enough jobs to go
around. To me unemployment is evidence that it is hard to FIND a job,
not that there are too few jobs. Lots of women fail to find a
husband, but you don't hear governments talking about man-creation or
a shortage of men.
For each individual to find a good job, society as a whole must solve
a very difficult combinatorial optimization problem, a bipartite
matching or assignment problem. Not an impossible problem, but we
certainly won't solve it as long as we ignore the combinatorial problem
altogether and try to do job-creation.
So, there you have it -- after complaining about Jay Hanson's
mistreatment of economists I go on to criticize them myself. But,
people, please, it's not personal, and it's not a prejudice, I just
think the universities need to add a few graph theory and computer
science courses to their economics curriculum.
dpw
Douglas P. Wilson [EMAIL PROTECTED]
http://www.island.net/~dpwilson/index.html
