Evaluating Conversations
The Loebner Prize Competition in Artificial Intelligence was
established in 1990 by Hugh Loebner and the Cambridge (Massachusetts)
Center for BehavioralStudies. It is awarded annually to the designer of
the computer system that best succeeds in passing a variant of the
Turing Test. In 1997 2,000 dollars and a bronze medal was awarded to
David Levy, designer of the Most Human Computer as rated by a panel of
5 judges, for his program Converse. Second place went to Jason Hutchens
and Bruce Cooper for SEPO. Third and fourth places were tied (based
upon the rankscores assigned by the judges) and shared by Richard
Gibbons (Julie) and Robby Glen Garner (Barry Defacto). Fifth was
Robert E. Medeksza, with BOB OS.
See http://www.loebner.net/Prizef/loebner-prize.html
for more background information.
Your task is to look at the transcripts on the conversations between
various judges and system, and comment on them. Transcripts of
conversations
generated by two of the programs: SEPO and Barry Defacto are given
here:
Extract of
conversations between judges and Jason Hutchens and Bruce Cooper's:
SEPO
Extract of
conversations between judges and Robby
Glen Garner's: Barry Defacto
(note these are abstracts from the original text).
A. For each program, select three sections of dialogue that illustrate
good examples of human-like conversation between the judge and the
program. Justify your selection by saying for each what the program did
well and why you selected this example.
B. For
each program, select three sections of dialogue that illustrate
examples of where the programs fail to produce good human-like
conversation between the judge and the program. Justify your selection
by saying for each where the program failed and why you chose this
example.
C. Imagine you were a judge. Give five questions you might ask and
explain why you think these would be good questions. Speculate on how
you think the programs might answer these questions.
D. Compare the two programs. What are their relative strengths and
weaknesses? Illustrate your answer with reference to the examples given
above. Which would you rate as best (you may have to say what you mean
by "best" to answer this).
1. Elbot: developed by Artificial
Solutions over the last 7 years. "It is our most
intelligent bot with an enormous knowledge base, and he is currently
used for advanced testing and sets the basis for the personality
module. If you want to have a chat with Elbot, you can visit him at http://www.elbot.com/."
2. Hal, Alan and the nursery....
Ai Research
is "a leading artificial intelligence research project. ......
Our expanding web site is an essential part of the emerging global
discussion about artificial intelligence. On this website, we showcase
the state of the art in pattern-matching conversational machines,
demonstrated by Alan,
and in reinforcement learning algorithms, demonstrated by HAL.
" "Ai is
hosting a
collection of 'Virtual Children': a
group of HAL personalities that
were trained by users who have agreed to make them available to the general public. If you wish
to choose one of these Virtual Personalities to speak to, click on HAL
Personalities to view the list".
Informatics Forum, 10 Crichton Street, Edinburgh, EH8 9AB, Scotland, UK
Tel: +44 131 651 5661, Fax: +44 131 651 1426, E-mail: school-office@inf.ed.ac.uk Please contact our webadmin with any comments or corrections. Logging and Cookies Unless explicitly stated otherwise, all material is copyright © The University of Edinburgh |