Was introduced by zadeh in 1965 to handle uncertainty

This preview shows page 17 - 20 out of 90 pages.

was introduced by Zadeh in 1965 to handle uncertainty in fuzzy systems,and has much in common with probability.Although mathematicians atfirst considered it to be a defective theory, possibility theory actuallytackles a different problem. Fuzzy logic has been widely used by the17
The Future of Artificial IntelligenceJapanese in the design and construction of household appliances.”(Barber, Botti and Koehler, 2002)Research conducted “by Newell, Laird and Rosenbloom (SOAR) isthe best-known example of a general architecture for an AI system.”(Barber, Botti and Koehler, 2002) As such, a fundamental feature of the“general architecture is its capacity to include many different kinds ofdecision-making, from knowledge-based deliberation to reflex actionresponses.” (Barber, Botti and Koehler, 2002)The “new agentarchitectures aim to strike a balance between these two factors, reflexresponses, for situations in the which speed is of the essence, andknowledge based deliberations, where the agent has time to take moreinformation into consideration, for forward planning, for handling situationsin which there is no immediate response available and to propose betterresponses modified exclusively to the situation in hand.Architecturessuch as SOAR have precisely this structure. By means of compilationprocesses like explanation based learning, they convert declarativeinformation at a deliberative decision making level into more efficientrepresentations until the decision eventually becomes a reflex action.”(Barber, Botti and Koehler, 2002)It is important to note that “agents in real environments need tohave the means of controlling their own deliberations and also be capable18
The Future of Artificial Intelligenceof using the time allowed for reasoning to perform the calculations, whichwill provide the best results.” (Barber, Botti and Koehler, 2002)Whenthese AI systems are then incorporated into more complex domains, theproblems that occur in real time will then become a problem for the agentsince this agent will not have the allotted time to find a reasonable solutionto the problem.Upon realizing the immense need to ensure that working methodswill continue to encapsulate more generalized decision making situationsand in recent years there have been two “potential techniques haveappeared, anytime algorithms and decision theory techniques.” (Barber,Botti and Koehler, 2002) The last building block of any agent architectureis the learning functionality and through the use of this “learning,reinforcement learning and compilation learning mechanisms can be usedfor all agents’ learning activities.” (Barber, Botti and Koehler, 2002) Theimportance of said mechanisms will actually depend on “what type ofrepresentation has been chosen which will be based on logic, and neuraland probabilistic networks are well known and much studied formalismsfor which there are a great variety of learning methods.As newrepresentations are created, such as first order probabilistic logics, it willbe necessary to create new learning algorithms for them.” (Barber, Bottiand Koehler, 2002)19

Upload your study docs or become a

Course Hero member to access this document

Upload your study docs or become a

Course Hero member to access this document

End of preview. Want to read all 90 pages?

Upload your study docs or become a

Course Hero member to access this document

Term
Winter
Professor
Mario Gomez
Tags
Artificial Intelligence, Future of Artificial Intelligence

  • Left Quote Icon

    Student Picture

  • Left Quote Icon

    Student Picture

  • Left Quote Icon

    Student Picture