Meeting notice: The 07-06-99 meeting will be held at 7:30 
p.m. at the Royal East (782 Main St., Cambridge), a block 
down from the corner of Main St. and Mass Ave. If you're new 
and can't recognize us, ask the manager.  He'll probably 
know where we are. 

Note:  There is of course no meeting on 06-29-99.  I am
sending this announcement out early because I will
be travelling for the next ten days.

Suggested topic(s):

AI and transaction rates: 

For the last four decades people have been speculating about the 
most plausible incubator for AI, defined as the ability to 
recognize (or better, learn to recognize) a variety of subtle 
patterns along a range of sensory modalities.  Candidates have 
included robots (on the theory that real world interactions 
require common sense and CS depends on AI), games such as chess 
and Go, and agents, especially mobile agents.   

The question is of interest to spectators in the Nanotech races, 
since NT would seem to require a pretty advanced level of AI to 
get any serious distance off the ground.  I am beginning to 
wonder if the most plausible testbed to watch is transaction 
processing, with transactions being defined as information sent 
and received from one machine or network to another (financial 
transactions are just a subset of these).  

Transaction rates go up with user populations, the complexity of 
the average user interaction, and the number and complexity of 
the applications being supported.   When all four of these 
measures are growing simultaneously, as they can in an expanding 
company competing actively in an ecommerce "space", aggregate 
network transaction levels look more like an explosion than a 
normal growth curve.   I recently had a conversation with an IT 
guy who figured that he had to assume that the transaction load 
on his company's network was going to increase 64x every twelve 
months for at least a couple of years and maybe even longer.  
While that is extreme, network managers routinely face 
infrastructure scaling issues that would have left earlier 
generations open-mouthed.

The relevant feature of transaction scaling is that it seldom  
happens in isolation; a rising hit rate on a retail website can 
force the pace of decision-making throughout the organization, 
from marketing to accounting.  This is turn raises the incentives 
to automate more and more of the functions that define middle 
management, both to ramp up decision rates and so managers can 
turn their attention to the larger questions of reorganizing 
their departments, which scaling also forces them to do, and over 
and over at that.  There are now marketing programs that define 
'target populations' for marketing campaigns automatically 
(Verbind) and I have spoken with people building network 
management simulators that forecast potential bottlenecks and 
prescribe remedies, and MRP software that monitor online auctions 
for a company and then auction off old inventory or production 
equipment when the price/cost equations are favorable.  These 
were all once archetypical middle management responsibilities.

If and when transaction increases grow even more vertical, these 
programs might start dealing with problems that now characterize 
senior management, perhaps eventually including strategic 
planning and mergers and acquisitions and the like.   Decision 
software might start figuring when to buy more decision software, 
which would allow natural selection to assert itself, with 
results that are not simple to predict.  Somewhere along the line 
of this development, some genuinely transferable AI principles 
might appear.  Given the amount of investor interest, which at 
the moment far exceeds the interest in robots, games, or agents, 
this domain seems as favorable to the enterprise as any other.

<><><><><><><><><><><><><><><><><><>

Announcements from Foresight:


	1999 Feynman Prizes in Nanotechnology

Two prizes in the amount of $5,000 each will be awarded to the 
researchers whose recent work has most advanced the development 
of molecular nanotechnology. Separate prizes are awarded for 
theoretical work and for experimental work. The prizes will be 
given at the Seventh Foresight Conference on Molecular 
Nanotechnology, October 15-17, 1999: 
http://www.foresight.org/conference/MNT7. 
http://www.foresight.org/FI/1999Feynman.html


	7th Foresight Conference on Molecular Nanotechnology

http://www.foresight.org/Conferences/MNT7
Westin Hotel in Santa Clara, California
October 15-17, 1999

Tutorial on Foundations of Nanotechnology
October 14, 1999
http://www.foresight.org/conference/MNT7/Tutorial.html


<><><><><><><><><><><><><><><><><><>

Announcement Archive:
http://world.std.com/~fhapgood/nsgpage.html.

<><><><><><><><><><><><><><><><><><>

Online discussion should be sent to 'nsg-d@world.std.com.  
Note: you must be subscribed to nsg-d to post to it and you 
must post from the subscribed address.   (An anti-spam 
thing.)  You can subscribe by sending the string 'subscribe 
nsg-d'  or 'subscribe nsg-d <address>' to 
'majordomo@world.std.com'.

If you have tried and failed to unsubscribe send the string 
'help' to majordomo@world.std.com.  If that doesn't in fact help 
see below:

Comments, petitions, and suggestions re list management to: 
hapgood@pobox.com