Re: poly: Econ, the final frontier...

From: Perry E. Metzger <perry@piermont.com>
Date: Wed Jan 07 1998 - 12:29:12 PST

Robin Hanson writes:
> Having been both an AI researcher and a social scientist, I suggest
> you reconsider your early impressions. Not only does AI have lots to
> offer (amid lots of crap), but technical social science has even more.
>
> I don't know where you got your impressions,

Well, first, AI:

I was a CS major and graduate student. I started out when I got to
school very excited by A.I. -- I imagined that AI researchers would be
the "best and brightest" and AI research would be exciting. Hundreds
of mind-numbing class hours later, and after lots of work with live AI
researchers, I gave up on AI. The topic *is* interesting, and *is*
capable of being the subject of real breakthroughs, but with some
small exceptions (the work the vision people were doing and some other
work I saw) I was pretty much disgusted with what I found. As just one
example, I saw NLP researchers re-inventing compiler parsers, only
worse, because they didn't know what they were doing even in that
regard. The "expert systems" crowd spent a whole lot of time
pretending they weren't just doing if/then trees, too. The sort of
basic research and scientific results one would have expected from the
field seems to have been totally lacking.

As for social science:

The small amount of "serious" social science I got exposed to in
school didn't seem like it qualified as science. Non-falsifiable
hypotheses, no controls, lots of cargo cult science. I must admit that
I didn't spend the amount of time looking at the topic as I did for AI
-- it was hardly the view of a professional from a nearby field with
extensive exposure. However, what I did see made me cringe.

> 1. The place to mostly reliably find the good stuff among the crap is
> in the top academic journals. It's *not* in popular articles on the
> subject, and *not* by taking a random class at a random school.

Well, perhaps, but random engineering classes and classes on compiler
design don't seem nearly as speculative and weak as random AI
classes. My random physics classes and even philosophy classes never
seemed as fuzzy as a random social science class. I admit that I
perhaps missed the cream of the crop, and I have no doubt that both
disciplines are important and have a great deal of potential -- but I
do, I must admit, have a "postjudice" about them -- that is, a
negative opinion based on exposure.

> 2. Undergraduate courses in computer science and especially economics
> typically show lots of distain for their students.

In the AI case, I was exposed to graduate level courses and
lived/worked with the graduate students and professors for years. I
must admit that in many instances my fellow systems and other
subspecialists in C.S. developed an amused and patronizing attitude
towards them, having discovered that the people researching how to
make computers behave "intelligently" were not exactly experts on how
to get themselves to work intelligently...

BTW, my undergraduate economics classes *were* pretty interesting, and
never had the soft feel that most subjects had. Then again, my first
econ class was taught by a very postjudiced and jaundiced escapee from
the Soviet block...

Perry
Received on Wed Jan 7 20:19:21 1998

This archive was generated by hypermail 2.1.8 : Tue Mar 07 2006 - 14:45:29 PST