|
|
This is only a preview of the paper
Click here to register and get the full text.
Existing members click here to login
|
|
|
Marvin Minsky Interview May 5, 1994
Interviewer: Douglas Riecken
Title: Agents and Agencies
D: It appears that the word agent is becoming popular both in the press and in industry. Personally, I have a strong interest with the idea of agency where many "simpler" things, call them agents if you will, "come together" and whose combined actions enable expressive and knowledgeable behavior -- the way that your Society of Mind theory (SOM - Minsky 1986) describes human intelligence. I guess I see an assistant as an intelligent machine composed of many agencies of "agents," not an agent being an intelligent machine. ...
5. ... In the present day jargon of computer interface agents, the word is used for a system that can serve as a go-between, because of possessing some specialized skill. ...
Agents and Agencies ====================
D: With that being the case, should we think instead of Phil as not merely a single agent, but as some large sort of agency? ...
So, in my book, I used the term agency to suggest the image of an office or an organization that is composed of several interacting sub-agents. ...
Another problem was that the English word agency already has the sense of an organization of agents, but many other languages have no such convenient pairs of similar words. ... What if our "agents" eventually reach a point where they believe they can do better with their own colonies as opposed to dealing with humans. ... We want them (agents) to help us do the things were not good at, but not get to good at doing those things. ... See (1) Marvin Minsky, "Music, Mind, and Meaning," Computer Music Journal, Fall 1981, Vol. 5, Number 3. An earlier version is in "The Neuropsychology of Music," (Manfred Clynes, ed) Plenum, New York, 1981 and (2) Marvin Minsky, "Jokes and their Relation to the Cognitive Unconscious," in Cognitive Constraints on Communication," L. ...
In the early 1980s it became clear (especially to Douglas Lenat, who was conceiving of his CYC project) that to make much further progress we were going to need a conceptual database adequate for understanding the most usual things that happen in the world. ... This led to a variety of ideas that included trying to get intelligent behavior from populations of essentially autonomous agents, trying to solve problems by situated-action methods (that is, by essentially reinitializing the machine after each action), or building essentially empty machines equipped with very general evolutionary capabilities. ... They must have a great deal of common sense if we are going to give those "agents" the authority to work on our behalf and spend our resources and money. ... Douglas Lenat is pioneering one approach -- of representing as much common sense knowledge as he can in a system based on logical set-relationships and frame based default-rich representations.
Approximate Word count = 5249
Approximate Pages = 21
(250 words per page double spaced)
|
|
|
|
|
|