COMMENT ⊗   VALID 00004 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	Conversation with John Kunz  7-Oct-82
C00005 00003	∂27 Nov 1982 1704-EST	CARBONELL at CMU-CS-C	progress?
C00010 00004	Questions for CPP
C00011 ENDMK
C⊗;
Conversation with John Kunz  7-Oct-82

Combining AI approaches with Mathematical models

Valdine, "Renal Physiology: ... Normal ..." [?1973?]
327-3680
  326-0681 stacey's

<<wrt my task:>>
Endocrine is TOO hard.  Let's start with something much simpler:
bodily fluids...
"The kidney system is like an electrical circuit in that both are circuits."

That is a functional specification, with certain structural considerations
	e.g. the topology is relevant,

working on circulating bodily fluids -- water, lymph, blood, ...

		
Body water → Kidney → Urine 
  ↑		↓
  |	     (thru <Pump>)
   \   ←     ← /

If there is no urine output, where might the problems lie?
Perhaps
(1) the Kidney → Urine passage is blocked, or
(3) the body-water → kidney passage, or
(2) the "<pump> vs urine" switch is flipped the wrong way, or
(4) there is no body water to begin with,
[or, a secondary consideration]
(5) the <pump> is no powerful enough to drive this...

----
John's model knows about connectivity, and throughput of this material.

Could it be twiddled to handle Heart (wrt blood flow) stuff?
Or be based on circuit analysis?
	[Here ask Dart people for their model of circuits...]

Goal is a system which can perform some task in this medical domain,
based on the knowledge base associated with that earlier (electrical) domain.

[ie. when trouble shooting: if there is no current, then that must mean ...
 or to teach, based on qualitative effect.]

Class of objects:
Holders,
  Batteries, Resouviours, ...
  Capacitors
Switch
   Filters
Pipes
   Wires
"Resistors"

Logic units...
∂27 Nov 1982 1704-EST	CARBONELL at CMU-CS-C	progress?
To: greiner at SU-SCORE

Russ,
 Have you made any more headway in your thesis project?  If you have any
new papers send copies my way -- especially if you managed to get your
ideas in a form concrete enough to incorporate in running programs.
 Good luck & forge ahead.
--JC
-------

∂27 Nov 1982 1443-PST	<CSD.GREINER at SU-SCORE>	Re: progress? -- One step forward, two steps back...
To: CARBONELL at CMU-CS-C
cc: CSD.GREINER at SU-SCORE

Jaime -

	Good to hear from you again!  Some ideas are beginning to take shape,
while others remain as fuzzy as ever.  In terms of general ideas:
I yakked with Pat Winston recently, and (based on that discussion) think
his most recent work on "form follows function" (my interpretation) will be
well worth reading.  This is in collaboration with Tom Binford (head of 
Stanford's robotics lab) and Mike Lowry (one of Tom's students).

WRT (my view of) analogy in general:
it's currently Mike (Genesereth)'s iteration of rewriting the 
"What's an Analogy Like" paper he and I are (slowly, slowly) producing.
(This will be a formal discussion of analogy -- trying to define what
this term means, and how to use it, in concrete terms.)
I will certainly send you a copy when it reaching a circulation stage,
and look forward to hearing your critique
(and hope that occurs in the not-too-distant future...)

	Towards implementing a running system, I've made some progress --
mostly in terms of deciding on an appropriate task to pursue, and the
relevant/optimal domains to consider.  Remember the analogies we all
heard when learning about electricity -- that one could think of an
electrical circuit in terms of water flowing through pipes, where batteries
were like resoviors(sp), and wires were like pipes, etc.?
Basically, I'm going to (try to) use these same analogies to build a new ES, 
capable of "understanding" electrical systems, based on an existing ES which 
deals with water flow situations.

The starting system may be taken from John Kunz's renal physiology system,
and the goal will probably be a QUAL-like explanation system 
(see [deKleer79]).
[Note: rather than build this destination system from scratch,
I might instead start with a weak version, ignorant of certain features,
and use these analogies to remove such deficiencies.]

The plan is to use the same basic mechanism which worked for the more
superficial
	"Viral Meningitis is like Bacterial Meningitis, 
	 except it is caused by a virus"
type analogies I mentioned a while back.  If all above works, I'll look into
other cases of this "deep analogizing" --
to build an endocrine understanding system based on an ES conversant in
digital circuits, or to understand something like the Actor formalism,
based on real world actors, or ...

----
Enough of this hand-waving.  I'm eager to hear what new and interesting
insights you have had of late, as well as the latest "gossip" about CMU
in general, and wrt to your tasks in particular.

Caio,
	Russ
-------

Questions for CPP

Current status of DART
	Where is code, and in what state?
	What assumptions were made?

System for Analog circuits?

Other relevant research

deKleer/Sussman
sophie  - js brown
"Watson" - at MIT (? brown)