![[Home]](/ailab/people/andel/pics/AIWiki.png)
Pages about agents in AIWiki:
here’s an interesting comment i came across (from MeatBall:StupidAgent?). The author is trying to define an “agent”, i guess (versus this term called a StupidAgent? which someone else invented):
“The computer works well for event-driven or scheduled events, but less so if working on its
own timetable. An agent, by definition, is something that acts or has authority to act. In an event-driven or scheduled system, the program reacts to the event or to the time prompting, which is an event, actually. So a StupidAgent? isn’t an agent in this sense, as it reacts, not acts. “
in other words, an agent is something that initiates/carrys on its own actions, rather than its motion being generally prompted by environmental cues. what do you think of this idea?
To suggest such a thing
necessitates the prerequisite existence of a soul or at least some sort of a priori cognition to reach agenthood. Does he think that humans are agents? How would he know?? His definition seems to run up against some signficant ontological problems, though I’ve only read the comment that you forwarded.
i don’t think the “acts without stimuli” thing is meant quite so literally; for instance, one could say that a word processor or even a cron job that does something like copying a given set of files is purely “reactive”, whereas a reinforced learner is “active” because it just tries stuff, and the old MSOffice? paperclip is “active” because it decided when to pop up rather than you telling it. both of them are still technically acting on prior instructions given at compile time, but you can see the sort of distinction i’m getting at.
Quite; I’d suggest that an agent could be defined in general as something which exerts force or influence, and in the context of our AI-type agents, something which does so more or less autonomously. You might type the command that launches it and gives it a goal, but it goes off and does its own thing until it’s finished. Seem reasonable?
– RobRix
So a batch file that trundles around looking for files that need to be archived, and archiving them, is an agent ??
– GL7
Actually, I do think of such background processes as primitive agents. – BayleShanks
Soul is not required at all, even humans are event-driven. For instance, one start acting when his level of boredom reach a certain level (event). – Juky
I don’t know if I would state so categorically that humans are event-driven, but that’s something for another time. I think it’s important to segregate the types of stimuli, though; boredom or hunger as a drive is very different from being poked with something sharp, which is in turn very different from receiving a command which says “do this absolutely now.” A command-line program like ftp is not autonomous; it does what you tell it to when you tell it to do it– or more often, when I’m (ab)using it, it fails to– rather than getting bored and wandering off under its own power.
Human beings build up mental models of the universe– and, I believe, primates do also– which we wander around in. Our meanderings in our representative-yet-imaginary universe provide a lot of the drive for our real-world actions. Trains of thought can lead to actions as easily as circumstantial events can, and if you think long enough on the representative bit, they start looking like actions themselves. It’s all very interesting.
As a definition of agent, I stand by what I wrote earlier– something autonomous which exerts influence. It doesn’t include our command-line ftp program, but neither does it include research projects which may exhibit some “intelligence” but at the cost of the modicum of control we desire from agents. Because honestly, I don’t think we want our agents getting bored and playing ball instead of researching or whatever it is they’re supposed to do… but neither do we want to have to do all the work ourselves.
– RobRix