From: James Rogers (jamesr@best.com)
Date: Fri Sep 15 2000 - 06:33:03 MDT
On Fri, 15 Sep 2000, Daniel Ust wrote:
>
> I agree. From a business perspective, too, the applications are arcane.
> Yes, certain forms of genetic programming, such as genetic algorithmns (GA)
> have been used in certain areas of engineering design where things are
> highly formalizable -- you can create a system of [nonlinear] differential
> equations to model, say, propellers blades, then use GA to solve them --
> there isn't much else. At least, there's little that someone is building a
> viable business plan around. No one, to my knowledge, is using them for,
> say, e-commerce or web development.:)
I am not sure that GA is actually useful for most classes of business
problems, or at least, I think there are methods that give nearly as good
results in a more deterministic fashion under the constraints of what
businesses want. While GA sounds good for offline, non-realtime problem
solving domains, it has never seemed well-suited for environments where
one needs to adapt to rapidly changing conditions in realtime. Also,
businesses are really uncomfortable replacing humans with machines in
these spaces if the business algorithms cannot be explained in excruciating
detail, which they can with people and plain old software (POS :^).
For example, I actually write (and sell) domain-specific commercial
engines that are designed to replace (and improve on) the adaptive pattern
matching ability of hundreds of people for realtime transactional systems.
While not "genetic", the engines actually have the capability to write,
rewrite, and test in realtime (can't have broken or flawed code hosing
mission critical online transactional systems), significant sections of
their own code based largely on fundamental realtime changes in the data
(and data structures -- it has to handle new data systems and
substructures that weren't even conceived of when the systems was written
while online, without human intervention) and patterns of usage.
I am currently working with an extremely large multinational to move an
entire division to one of these engines, but it is important to note where
the real business resistance has come from. While the software can
demonstrably respond much faster without making mistakes than human
operators using POS, and make modest but significant improvements in
direct profits through superior handling of anomalous situations, the
executive management was very concerned with the fact that it is not
easily possible to figure out why and how a specific transaction was
handled in the way that it was. Also, some of the code written by the
system is nigh inscrutable (never mind that it can change at any time), so
reverse engineering it is a major pain in the ass. In short, while the
software is good at what it does, they have difficulty trusting software
that can modify itself to do the job better based on moment to moment
market conditions (as though human software was doing so much better).
This is understandable though; a flaw in the software can contractually
bind them to lose millions of dollars at computer speeds, whereas flawed
humans tend to propagate damage at a much slower rate.
So, in my opinion, the reasons you don't see genetic algorithms in most
business systems are two-fold: 1) most business systems are transactional
online systems that aren't particularly well suited for GA, and 2)
business executives are leary of online transactional software that isn't
based on static algorithms.
-James Rogers
jamesr@best.com
This archive was generated by hypermail 2.1.5 : Fri Nov 01 2002 - 15:31:00 MST