[net.ai] AI Thesis Poll Response Summary

marie@harvard.ARPA (Marie desJardins) (01/04/85)

Here are the results of the thesis poll I posted last month:

11 people responded to the poll.

They were geographically distributed as follows:

	2 in Maryland (2 at UMCP)
	1 in Texas (UT Austin)
	3 in Canada (Alberta, Newfoundland, Calgary)
	3 in New England (2 at Rochester, 1 at MIT)
	2 in California (UCLA, UC Davis)
	1 Elsewhere (VU Information, Amsterdam)

I tried to categorize the topics, but some of them seemed to defy
categorization.  When people responded with more than one topic, I split
up their "vote" (i.e. if someone gave 2 possible topics, each category
has 1/2 "vote").  I lumped a lot of stuff into expert systems that
probably didn't belong there.  Below are all the individual topics in 
condensed form if you'd like to see what the responders really said.

	Natural language	2 5/6
	Expert systems		1 1/2
	Learning		1 1/3
	Robotics		1 
	Vision			1
	Planning		1
	Search			1/3

Actual topics:

- natural language understanding -- modeling what humans *might* do -- using 
	a parallel activation net.
- self-organizing retrieval system for graphs. 
- knowledge-based image understanding.  
- Most likely picks: Automated theorem proving; knowledge based (expert) 
	systems, or; Natural language processing.
- "Implementing Algorithms for Segmentation of Aerial Images on Parallel
	Processors" 
- collaborative systems (robots, etc. who on basis of equality try to 
	optimize their overall behaviour in a non-predictable environment).
- knowledge acquisition, particularly under the paradigm of "learning by 
	asking questions."
- natural language processing, or expert selection and linkage.  
- knowledge representation, learning, or search strategies applied to 
	computer chess.
- Planning. I am going to use temporal constraints to deal with resources 
	and goal subsumption. 
- My MSc thesis topic is graphical tracing/debugging of Prolog.
- "optimal prediction and scheduling of work in a medium-term operating 
	system scheduler."  it is an application of
- classical pattern recognition to design of scheduler's for operating
	systems.  
- reference in Nat. Lang.


Thanks to all who responded!

	Marie desJardins
	marie@harvard