Meeting notice: The 01.07.03 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.  More details below.

<-><-><-><-><-><-><-><-><-><-><-><-><-><-><-><-><-><-> 

Suggested topic: Massively Distributed Artificial Intelligence

AI can be sorted into three levels:  recognition of primitives, recognition 
of wholes,  and the retrieval or inference of properties associated with 
those wholes.  Examples might be recognizing edges or planes (or phonemes); 
connecting (or seperating) those edges in order to reveal a table or a 
horse, and retrieving or inferring the properties relevant to a potential 
interaction with that table or horse. 

Of these the third task is the most important, in that no subtype of AI can 
possibly get far with skills in this department.  One approach to the 
problem is Doug Lenat's CYC project in Austin, in which a fixed number of 
smart programmers work for decades on building a properties database.  This 
approach has the defect that every real object has an infinite number of 
properties that cannot be calculated from a subset of its other properties. 
There is no way of generating the price of a product, or its history or 
function (often), from its physical properties, and vice versa. Thus a 
serious AI program has to know what it is ignorant about, be able to search 
for the appropriate value, recognize and retrieve it if it exists, or come 
up with a WAG if it does not.  It has to be learning constantly.  The model 
of depending on a small number of supersmart coders to hammer the 
properties in one by one does not seem to scale to the requirements of the 
application.   

What would appear to be needed is a standard ontology format that accepts a 
very wide range of knowledge types.  For instance, an automation tool, on 
being told that a given story was set in 1957, should be able to send out a 
call for a number of cultural objects -- cars, houses, fashions, makeup, 
dances, conversational tropes, body language -- appropriate to that time. 
To pick a topic closer to our interest, when we tell an assembler to design 
a material we would want it to know the latest science, the best prices, 
the most interesting application possibilities for a material of that sort, 
and so on. 

It is hard for me to see how the depth of information envisaged can be 
reached unless everyone uses the format routinely for their own purposes 
and then makes their efforts generally available. (Essentially, everyone 
becomes a volunteer for CYC 2.0.)  The latter might be acceptable, but the 
former seems like a real speed bump, since it is hard to think of any 
formatting procedure that would do what we want it to do without being so 
inconvenient that it would drive people away.   

Thoughts? 




<-><-><-><-><-><-><-><-><-><-><-><-><-><-><-><-><-><-> 

In twenty years half the population of Europe will have visited the moon.

				-- Jules Verne, 1865

<-><-><-><-><-><-><-><-><-><-><-><-><-><-><-><-><-><-> 

Announcement Archive: 
http://www.pobox.com/~fhapgood/nsgpage.html.

<-><-><-><-><-><-><-><-><-><-><-><-><-><-><-><-><-><->

Legend:

"NSG" expands to Nanotechnology Study Group.  The Group meets on 
the first and third Tuesdays of each month at the above address, 
which refers to a restaurant located in Cambridge, Massachusetts.  

The NSG mailing list carries announcements of these meetings and 
little else.  If you wish to subscribe to this list (perhaps 
having received a sample via a forward) send the string 
'subscribe nsg'  to majordomo@world.std.com.  Unsubs follow the 
same model.  

Discussion should be sent to nsg- d@world.std.com, which must be 
subscribed to separately.  You must be subscribed to nsg-d to 
post to it and must post from the address from which you 
subscribed (An anti- spam thing).

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