Roskilde University Digital Archive >
RUCs Digitale Projektbibliotek / RU Digital Project Library  >
RUC Basisrapporter / RU Basic Studies Projects >
Naturvidenskabelige basisrapporter / Natural Science Basic Studies Projects >

Please use this identifier to cite or link to this item: http://rudar.ruc.dk/handle/1800/616

Title: Kunstig Intelligens
Other Titles: Artificial Intelligence
Authors: Andersen, Niels
Baysal, Derya
Bonde, Heidi
Hannesson, Kristófer
Holst, Anne-Marie
Advisor: Braüner, Torben
Issue Date: Jan-2004
Abstract: Dette projekt handler om kunstig intelligens. Vi tager udgangspunkt henholdsvis i det klassiske- og det konnektionistiske paradigme. Vi bruger orddeling som case, for at vise forskellen på paradigmerne, hvor det klassiske paradigme benytter symbolmanipulation, og det konnektionistiske benytter kunstige neurale netværk. Den menneskelige intelligens bliver defineret ud fra psykologi, mere specifikt kognitionsbegrebet. Turing definerer kunstig intelligens i forhold til en test han har lavet og ikke ud fra en psykologisk retning. Searle deler kunstig intelligens op i to kategorier, stærk og svag kunstig intelligens, hvor Searle mener, at der ikke kan opnås stærk kunstig intelligens, dvs. at et system kan opnå bevidsthed, men at man kan simulere intelligens, hvilket kaldes svag kunstig intelligens. Konklusion er, at det umidelbart ikke er muligt, ud fra orddelingseksemplet, at sige hvilke af paradigmerne der er bedst. Men vi kan se fordele ved både det klassiske- og det konnektionistiske paradigme. Det konnektionistiske paradigme er f.eks. godt til at generalisere, og kan, i orddelingseksemplet, klare ukendte ord lige så godt som kendte ord, hvis netværket er trænet nok. Det klassiske paradigme klarer orddelingen lige så godt som det konnektionistiske paradigme, hvis undtagelseslisten, der indeholder alle de irregulære ord, er lang nok. This report relates to artificial intelligence. We discuss the relationship between the classical and the connectionist paradigms, with word division as an example. The classical and the connectionist paradigms respectively use symbol manipulation and artificial neural networks in implementation. Human intelligence is defined in psychology, within the field of cognition. Turing defines artificial intelligence from a test he has created, and not from a psychological point of view. Searle divides artificial intelligence in two categories, strong and weak artificial intelligence. Searle believes that the classical paradigm cannot achieve strong artificial intelligence, because consciousness can not be achieved. However, artificial intelligence can be simulated, which is called weak artificial intelligence. The conclusion is that it is not possible at this point to tell which of the two paradigms is preferable in the word division case. We can see advantages in both paradigms. The connectionist paradigm, could divide words that it was not trained with. The classical paradigm is expected to perform equally well, compensated with a list of irregularities in word division.
URI: http://hdl.handle.net/1800/616
Subject: RUC projektrapport / Thesis; NAT Basis;
Appears in Collections:Naturvidenskabelige basisrapporter / Natural Science Basic Studies Projects

Files in This Item:

File Description SizeFormat
Kunstig Intelligens - Artificial Intelligence.pdf468,95 kBAdobe PDFView/Open


This item is protected by original copyright

Recommend this item

Items in RUDAR are protected by copyright, with all rights reserved, unless otherwise indicated.

 

Valid XHTML 1.0! RUDAR Software Copyright © 2002-2011  Duraspace