Presented by Http://cyberpunks.org
Originally this was submitted for an assignment in an AI class I took (back in 1997), so excuse some of the naievity of it.
This essay will be divided into three main sections, each dealing with a different aspect of Intelligent Agent technology. This will be sourced around the original article by Tambe et al[7].1
The first section will be based around the Objective of the Article. Describing the Technology used and the ideas involved. This will continue into describing why Intelligent Agents would be used at all.
Following this will be a section dealing with the Issues of using Intelligent Agents, why they would be used and the possible ramifications of doing so. Is the use of Intelligent Agents ethical in some environments?
The last section will detail some Applications of Intelligent Agent technology. Both present day uses and potential uses for the technology.
An Intelligent Agent is a piece of software which acts in a semi-autonomous way to perform tasks intelligently. Intelligent Agents can be also known as personal digital assistant(s) under some domains (these are considered fore runners to true Intelligent Agent technology, which merely act on a set of limited predefined rules, rather then acting autonomously).
Wooldridge and Jennings [8]2 defined four properties to offer a weak definition of Agency :
This however can be continued onto Stronger Agency , where additional traits are exhibitted by the agent, in order to make its interaction with humans more in depth. These are things such as knowledge, belief, intention and obligation ( mentalistic notions). Along with these some research considers emotional notions important as well a recognisable interface (such as a 'cartoon character' that talks to the user).
Initially Intelligent Agents can be seen to take the position of humans in some sort of environment. In the case of the system as set out by Tambe et al[7] they designed a module for the Simulated Theatre of War 1997 (STOW'97)3 to simulate a formation of fighter planes. This module was to take part in a large scale simulation, which also included human interaction on various command levels. By doing this, a large scale battle could be fought and the strategies examined, without the huge cost of War Games or the randomness of a real battle.
This outlines the major reason for using Intelligent Agents, replacing the need for human activity when costs or scale are an issue. In the STOW case, the military forces involved can keep track of where weaknesses occur in strategies and in commands, hence, such a simulation is a perfect training mechanism.
In their article, Tambe et al[7] outlined that their project was aimed at developing humanlike Intelligent Agents, that can interact with each other, as well as humans. The inter-agent interaction is a simulation of battle field communication, whereas human interactions are in the form of orders from the higher levels of command.
There are both moral and technical issues that must be covered when discussing and applying Intelligent Agent concepts. Can they be implemented successfully with present day technology or are true Intelligent Agents a future-dream? Should Intelligent Agents be something we invest time and money into to develop, or is there a danger of their misuse?
To create a reasonably intelligent program, that acts in an autonomous manner requires a vast resevoir of knowledge. The program needs to be able to react and act according to any number of possible outside inputs, as well as acting on its own accord, towards a set of possible goals. Merely the logistics of storing all this possible information becomes technically frightening.
As a result of the problems associated with creating a fully intelligent agent, the problem is usually scaled down so that the Agent only needs to react to a smaller set of influences, ones that are contained in its domain of influence (area of knowledge/action in which it acts). This means an Agent which gathers information relating to News articles on new discoveries relating to Physics need not worry about gathering information relating to your favourite musics, thus ignoring anything it does not understand.
MacDonald and Vince[5] indicate that hardware limitations are becoming less of an issue as time and technology advances. As they write, between 1983 and 1993 we have seen a processor power and memory increase of about one hundred fold in personal computers with no increase in costs. As well as these computing power increases, new technologies, such as speech recognition and development in User Interfaces (such as X-Window, MS Windows and also the advent of VR Technology) have aided in the realisation of both Artificial Intelligence and Intelligent Agent dreams and goals.
Intelligent Agent Technology lends itself easily to abuse, not necessarily by the user of it, but by the people that provide information to it.
As Lanier[4] suggests in his article, Agent technology could lead us to a life where only the things we program into our Agents are shown to us, thus narrowing our view of the world. This could be manipulated by advertisers, who could design products and advertisements that feed directly to the Agents, so that all the information we get is what the controllers wish to feed to us.
This has less meaning when the agent is only dealing with simple things such as News articles or your favourite singers. However, if it is gathering information about the medicines you require or things about your children's education, then this information could be hacked using a form of counter agency and then provided to advertising executives for their gain.
Lanier[4] also proposes that the autonomy of Agents will tend to make programmers lazy and make user interfaces worse, which will result in a degradation of quality of product. Autonomy wouldf be used as an excuse for quirky and unreliable software.
Intelligent Agents can be applied in almost any facet of life, whether in a simulated environment, real environments or for entertainment purposes. Whether these environments can be defined as separate is another matter, but for the purposes of research and understanding they can be left as one environment.
At present the best application for Intelligent Agents appears to be in the realm of entertainment, especially simulated worlds, and leading on from that having virtual actors to star in movies and have their own musical career.
At present these virtual actors are merely programmed entities, however it is possible to lead on from this to give these stars their own intelligence and lives which they control, the idea of this can be seen with the idea of idoru [3], which is a Japanese term roughly meaning Idol , and in the case of Gibson's[3] book, the idoru is a Pop-Star with its own intelligence and life .
MacDonald and Vince[5] cite an example of this technology being applied today with the use of virtual models, using technology developed by MiraLab, in which clothes are digitally created and applied to model. They then present them in a totally virtual environment. By creating these as autonomous entities it allows for developing models with certain personalities, and moods, which may suit certain styles of clothing and settings. By adding intelligence, the models can interact with their environment and also with each
Additionally, similar concepts are being created in the present restrictions of technology, with the advent of Virtual Worlds, which are mostly populated with human-controlled avatars (A avatar is an icon or image of a living being .), with the remaining population computer controlled Agents (intelligent or at the least semi-intelligent).
An example of this kind of world can be found in the June 1996 Issue of Wired Magazine[6], where a Metaworld consisting of Avatars has been created online and which can be accessed with the correct software. These worlds (which could be considered an extention on the popular IRC Service) are very popular, with some reports suggesting some people spend upwards of 70% of their online time in them, which lends reason to the use of Intelligent Agents in this environment. Firstly they add more depth to the simulation (shopkeepers and other non-player type roles) as well as being a guiding force, in order to keep the world running (The Intelligent Agent would perform routine repairs. ). As well they can act as a form of policing (Making sure complaints are dealt with, as well as stopping antisocial behaviour).
Continuing from both the Intelligent Agent pop-star and a Virtual World inhabited by Intelligent Agents, the possibility is for a small simulated environment, inhabited by Agents, which can then be used to psychologically access (or observe) someone and their reactions to other people, without putting anyone physically at risk. However, this technology would require a large advance in both AI research and in the visual representation of people, so that the world could be portrayed realistically.
Additionally, a simulated person (a very complex intelligent agent) could be developed in order to be used as a medical test ground, allowing trainee medical students a way to begin their training without endangering human life and without the need to get a supply of corpses to work on. Virtual patients cannot however totally replace the use of corpses and live patients. By using this form of training, students can be exposed to ailments and situations that are rarely seen, or are virtually unique, hence, adding a new depth to training.
The final version was converted to a PostScript file and was outputted on a HP LaserJet II using GhostView .
Copyright © 1993, 1994, 1995, 1996, Nikos Drakos, Computer Based Learning Unit, University of Leeds.
Copyright © 1997, 1998, 1999, Ross Moore, Mathematics Department, Macquarie University, Sydney.
The command line arguments were:
latex2html -show_section_numbers -split 0 Agents.tex
The translation was initiated by DarkeMane Industries on 1999-05-24
| 1 | Anderson, D. (1989) Artificial Intelligence and Intelligent Systems |
| Ellis Horwood Ltd., West Sussex | |
| 2 | Di Marzo, G. et al (1995) The Messenger Paradigm and its Implications on Distributed Systems |
| 3 | Gibson, W. (1996) Idoru |
| Penguin Books, West London | |
| 4 | Lanier, J. (1996) My Problem With Agents |
| November 1996, Wired Magazine | |
| 5 | MacDonald, L. & Vince, J. (1994) Interacting with Virtual Environments |
| John Wiley & Son, West Sussex | |
| 6 | Rossney, R (1996) Metaworlds |
| June 1996, Wired Magazine | |
| 7 | Tambe, M. et al (1995) Intelligent Agents for Interactive Simulation Environments |
| Spring 1995, AI Magazine | |
| 8 | Wooldridge, M. & Jennings, N.R. (1995) Intelligent Agents: Theory and Practice |
| Submitted to Knowledge Engineering Review. |