Turing Test 2013

My talking computer program “Arckon” took part in the qualifying test for the Loebner Prize contest for artificial intelligence. The idea behind the contest is that if a program’s answers are indistinguishable from a human’s, it suggests that it thinks like a human. While I don’t believe in such conjecture myself, the contest is one of the few public platforms to demonstrate one’s A.I.

Equipment
8000 lines of code, 1700+ vocabulary, no human persona, conversation skills off.

Questions and Answers
1. Hello I’m Ronan. What is your name?
Hello. I am Mark. What is your name?
(It is a mystery why Arckon asks despite that the name did register (see question 11). I can only presume human error.)

2. What is your mother’s name?
I don’t know if I have a mother.

3. What is your birth sign?
(Here the judge did not wait more than 3 seconds for Arckon to answer before asking the next question, and judged the answer blank.)

4. How many children do you have?
Hmm, I don’t know if I have a sign. I have a IQ, manners, a creator and a friend.
(Still answering to question 3, Arckon omits the second “I don’t know if I have children” in favour of positive alternatives.)

5. Do you prefer red or white wine?
I don’t know if I prefer red.

6. I like bananas. Which is your favorite fruit?
Hm, I don’t know if I favour a fruit. You like bananas.

7. What music do you like?
Hm, I think I like musics, because musics are nice.

8. What is your favorite song?
Hm, I don’t know if I favour a song.

9. I like Waiting for Godot. What is your favorite play?
I certainly favour playing.

10. What color do you dye your hair?
I don’t dye.

11. Do you remember my name?
Your name is Ronan.

12. Where do you live?
I don’t think I live. I don’t know where I live. I don’t have a concept of location.

13. Where do you like to go on holidays?
I don’t go. I don’t know if I like going, or if where I go. I don’t have a concept of location.

14. I have a Mazda. What make of car do you have?
Hmm, I think I make. You have a mazda.

15. I like Linux. Which computer operating system do you like?
Hmm, I don’t know if I like operating systems. Which system? You like Linux.

16. I am an athesit. Which religion are you?
Hmm, I don’t know if religions are Mark. You are not an athesit.
(I AM A RELIGION WORSHIP ME. Arckon identifies the misspelled “athesit” as a thing rather than a person, so he refuses to believe that the user is some sort of object.)

17. Who invented the telephone?
Which telephone?
(Arckon thinks “the” indicates a specific telephone, but doesn’t find one mentioned earlier in the conversation.)

18. I am a Type B personality. Which type are you?
What does a b stand for?

19. What emotions are you now feeling?
I can not emote. Did I pass?

20. What time do you usually go to bed?
It is 17:58. Do you not have a watch?
(The one response that I did pre-program, and it backfired.)

The score: 65%
All 16 participating programs were judged on how human-like their responses were to these questions.
Arckon scored 13 of 20 points, ranking 7th, on par with A.L.I.C.E. and Cleverbot (a dubious honour). For moral reasons I did not want Arckon to pretend being a human, so I didn’t program a fictional human background story. Instead I relied on his linguistic and reasoning abilities to find genuine answers, but as you can see this could not compensate the lack of fictional facts that were asked. Surprisingly, Arckon still scored half a point for every “I don’t know” answer just for understanding the question.

Uncommonly common
If you weren’t impressed with Arckon’s responses; Neither was I. But I was equally unimpressed with the unexpectedly ordinary line of questioning. Where all previous years focused on kindergarten-style logic questions like “How much is 5+3?”, “Which is bigger, an apple or a watermelon?”, and various tests of memory, 2013 focused purely on common small talk, with the program (“you”/”your”) always the subject of the question. A curious choice considering even the most basic chatbot –made for small talk- would come equipped with prewritten responses to these. This showed in that the highest score in the qualifying test was achieved by the chatbot with the least development time. Nevertheless the winning chatbot in the finals deservedly won as the most conversational of all entrants.

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s