INTELLIGENT AGENTS Examples Internet filtering browsing 60 agents
INTELLIGENT AGENTS Examples ©Internet - filtering, browsing; 60 agents on http: //ai. ijs. si/mezi/agents. html e-commerce: http: //ai. ijs. si/Sandi/Intelligent. Agent. Reposit ory. html ©http: //www. csc. liv. ac. uk/~mjw/links/ ©PC - system agents, Office’ 97
Intelligent systems, intelligent agents © New AI directions: cognitive and applications © Advantages: adaptable, flexible, able to learn, userfriendly, “bluff” intelligence © A typical agent: insurance agent (M. Minsky); many users © Other types of agents: art. life, static-mobile, distributed, for people or computers © Intelligent nformation society
Truly intelligent? AI DEVELOPMENT, TECHNOLOGY
Bill Gates ©. . computer of the future - an intelligent computer assistant, a kind of secretary, capable of communicating and executing simple mundane tasks. The new system will be capable of talking, listening, seeing, and will have other anthropological features like faces capable of expressing gestures.
Properties of intelligent agents ©autonomy - ability to perform tasks and decisions without direct intervention of humans ©social ability, ability to interact with humans and agents ©responsiveness, the ability to perceive the environment and respond to changes
©proactiveness, the ability to take initiative and to exhibit goal-directed behavior ©adaptability, the ability of an agent to modify its behavior ©mobility, the ability to change physical location ©veracity, assumption of no false information ©rationality, ability to perform reasonably
Properties - Pattie Maes ©Observes a user ©Gets feed-back from a user ©Gets direct instructions from a user ©Gets experience from environment ©Agent and user communicate, control, execute ©Agent learns according to interests, wishes and desires of users
Properties - Etzioni © Autonomy when executing tasks; gets task descriptions from a user, modifies it, performs it on its own © Time continuous – work all the time © Personality - speak too much © Able to communicate © To adapt to each single user – personalization © Mobility
“Simple rules” © does it perform typical user-oriented functions (insurance agent) © autonomy - performs actions on its own (yes) - is prediction of actions possible (no) © adapts to each specific user - different reactions for the same error © works all the time, looks around (mobile) © data - information - knowledge
Properties - summary © General, not exact definitions - “ideal agent” – theoretical, nonexistent real agents only approximations with some properties - bordears soft, not exact © “agent” is a slightly different program © Similar relations: non/structured programming; modular/object; information systems/operation systems/expert systems; data/information/knowledge © some people don’t understand the difference
Types of agents - Etzioni ©Co-drivers – suggest where to go to ©Drivers – listen to suggestions by users ©Secretary-assistant, gets strategic directions and performs actions on its own
Slovenian agents © 1993 IOI, interface VAX/VMS; B. Hribovšek, M. Gams © 1996 EMA, an employment agent for Slovenia on Internet, M. Gams, A. Karalič, M. Grobelnik, M. Drobnič, P. Golob; National Employment Office © 1998 Personal Web. Watcher, D. Mladenič © 2000 Shi. Na, A. Pivk © 2000 Active. Tools, USA, A. Bezek
EMA - EMployment Agent
An Internet Agent
An Agent of Agents
Communication with anybody on the Net ©EMA observes a user communicating with an employment database through a form ©EMA repeats the query ©EMA modifies the query ©EMA “sucks” the database ©EMA reads any Internet employment database
CONCLUSION ©Intelligent agents are among the most prospective new SW breeds; even now world record holders ©Intelligent agents represents a marriage between AI, intelligent systems, and information society
- Slides: 17