Re: Special difficulties of AI terminology

From: Eliezer S. Yudkowsky (
Date: Tue Feb 19 2002 - 15:43:10 MST

Lee Daniel Crocker wrote:
> Why avoid neologism at all? When treating specific concepts within a
> field, and especially within a specific paper, neologizing in much
> preferred to overloading. Indeed, I might argue that creating new words
> for new concepts is part and parcel of the cretive process itself. The
> only thing that's bad is creating new words for old concepts that people
> already understand--crackpots do that to hide the fact that they aren't
> actually creating new ideas but just dressing up old ones. If your ideas
> are truly new and different (or even just more specific in an important
> way), then by all means makeup new terms. I wouldn't be able to teach
> you about poker strategy without using a bunch of terms like "outs" and
> "counterfeits" and "free cards" and "drawing dead". Why should I expect
> you to be able to explain AI to me without first defining some new
> terminology?

Well, my first reaction to this was to laugh and say, "I give up!" Some days
you just can't win...

On reflection, though, I think that there are two dynamics here. Anyone who
wants to think precisely will need to use precise terminology. So if you're a
bad writer and you want to think precisely, you use complete neologisms -
arbitrary, hard to remember or identify once you're a few pages out of
context. Good writing practice is to minimize this impact by adapting
existing terms where possible, and using recognizable word combinations where
a genuinely new term is needed. So it's preferable - in terms of being able
to think clearly and reason clearly - to use neologisms rather than not have
any new words at all, but it's also preferable to use recognizable language
rather than pure neologisms.

-- -- -- -- --
Eliezer S. Yudkowsky
Research Fellow, Singularity Institute for Artificial Intelligence

This archive was generated by hypermail 2.1.5 : Fri Nov 01 2002 - 13:37:40 MST