poly: Emerging Technology

At 08:08 PM 1/2/98 -0800, Tim May wrote:
>At 4:50 PM -0800 1/2/98, Anders Sandberg wrote:
>>One area I think will get BIG very soon is knowledge management
>>software of various kinds. We now have information, and we are getting
>>better at information management (still far to go there, but we are
>>advancing). The logical step is software to manage quality information
>
>Now I'm _really_ skeptical about his one. Not because it doesn't sound
>nice..."knowledge management" is like motherhood. But what's it mean?
>
>Is it data base management with some data mining thrown in? (I hope so, as
>I have a heavy position in Oracle.) Is it AI? (I hope not, as my AI
>companies have gone belly up in the last decade.) So, aside from hand
>waving, what is it?

I am pretty sure Anders is referring to a particular manifestation of a
lack of a technology. There is a key piece of technology that is becoming
increasingly necessary, but to date does not meaningfully exist. There is
no technology that transparently automates inter-domain (information
domains) correlation. The current technology for managing knowledge is
either cross-domain capable but painfully explicit and non-automated
(object/relational databases), or is domain constrained and very poor at
crossing domain boundaries but reasonably automated within its domain
(expert systems, current AI in general). This is one of the areas where
humans are currently irreplaceably in large information systems.

I have been doing research in this area for a few years now. I believe
that the technology (when it comes) won't be AI exactly, but will be
hovering somewhere on the border.

>Sure, if a killer app comes along it'll all be retrospectively obvious we
>should have invested in it. But the hard part is looking forward, not
>backward.

The is actually quite a few relatively obvious applications for this. It
is useful just about anywhere an individual has to sift through multiple
databases or broadly constrained information domains. For example, a tech
support database, a legal database/archive, as a general indexing tool for
very large data sources, or a database of research papers. The improvement
in useability for many applications would be enormous, but it is unclear
the extent of the ramifications since this capability does not exist yet.

Current indexing technologies are very primitive and remain fundamentally
unchanged since the first concordances were written by religious orders in
the dark ages. While we have found ways to vastly improve the indexing
technology (e.g. Inktomi), indexing is commonly used in software beyond the
scope of applicability because there is no alternative technology for
knowledge access.

>
>(Another of my contrarian views is that the rate of knowledge growth is
>_not_ accelerating, but is slowing down. Not that the number of bits and
>bytes of so-called knowledge is not growing exponentially...they are. But
>the discovery rate of major new facts, major new paradigms, etc. is
>slowing. Which is not at all surprising to me. I can elaborate in other
>posts.)

Actually, I am more or less in agreement with you here. The volume of
knowledge has gone way past the point where our existing knowledge access
technologies scale well. It is becoming increasingly difficult to
ascertain what we know and don't know. This is forcing people to spend an
increasing amount of time focusing "inward" on existing data spaces rather
than "outward" on new frontiers. There is no way to succinctly organize
everything you need to know regarding the current state-of-the-art when it
involves sifting through Terabytes of related knowledge, 99% of which is
useless to your research.

-James Rogers
 jamesr@best.com
Received on Sun Jan 4 08:10:07 1998