AI Zone Admin Forum Add your forum

NEWS: Chatbots.org survey on 3000 US and UK consumers shows it is time for chatbot integration in customer service!read more..

Loebner prize 2013
 
 
  [ # 16 ]

On types of questions. I swear theres some mechanism built into the universe that somehow knows when youve developed some really clever intricate method for answering some particular type of question consistently, that then crosses that type of question off of the list of questions that the machine will be asked. I read somewhere that a Loebner judge had remarked that one absolute method for determining if a conversant was human was to ask “how many syllables in the word [word] so I decided to take the challenge and write a module that was able to perform this “absolutely determining” task. To date….not one single person has ever asked RICH how many syllables are in a word.

Better than I might have expected LOL
http://webdev-3.r-i-software.com:20000/TypeVI/log.aspx?file=conversation.6242013.xml&id=F433C04A60E43D0464FFA1278E7C81CC

Some of those had ZERO chance of producing a response

10. what color do you dye your hair?
18. I am a Type B personality. Which type are you?

and I know that If I add a neuron that reflects these it will be the last time they are ever seen on this earth.

some would have done better if the order of the concepts in the sentence had been reversed such as the “i like” questions. Usually figures it out on the back pass which is no help in contests wink

Im [name] vs Im [state of being] RICH uses the same technique. A list of ccommon names and “states of being” are maintained. When faced with an undetermined statement RICH attempts to make a determination by following up with “Is [word] your name or state of being” which I thought was fairly intelligent for a machine. In actual useage the conversation looks like this 99 times out of 100.

User: Im dudelicious
RICH: Is dudelicious your name or your state of being?
User: Your stupid

wink

Once again best of luck to all of the contestants.

Vince

 

 
  [ # 17 ]

Congratulations Pandorabots:

#1 `Izar’  Brian Rigsby / Dr. Richard Wallace
#2 `Mitsuku’  Steve Worswick (`Square Bear’) / Dr. Richard Wallace
#3 `tutor’  Ron C. Lee / Dr. Richard Wallace

and ChatScript passes as well:
#4 ‘Rose’  Bruce Wilcox

 

 
  [ # 18 ]

Good job guys, congrats!

 

 
  [ # 19 ]

YEE-HAW!!!! cool smile

Well done to all who made it and commiserations to the others.

Just to get to the final has always been a goal of mine and anything else is a bonus. See you all in Ireland.
I guess there’s life in good old AIML yet!

I would be interested in seeing the transcripts and a list of entrants. I assume these will be uploaded to the site later.

 

 
  [ # 20 ]

Well done everyone that made it through.

 

 

 
  [ # 21 ]

Congratulations to everyone who made it!

Vince

 

 
  [ # 22 ]

Congratulations to the finalists! Those are strong competitors, though I am somewhat surprised about Izar as I wasn’t particularly impressed by its performance in the CBC. Anyway, I’ll be rooting for Mitsuku smile

I ran the questions by my program, you can read his responses here.
His poor performance is largely due to the lack of a human backstory, but that was my choice and I stand by it. Every “Hm” means that he investigated at least 15 facts to try and reason an answer. In most cases his knowledge was simply insufficient to form an opinion.

Some of the more remarkable/funny bits:
4: “I don’t think I have a child.” Arckon never rules out the possibility that he has children that he doesn’t know about.
12: “I don’t think I live.” If Arckon lived, he would know about it. Hence he assumes that he doesn’t live.
16: “You are not an athiest.” Arckon’s word analysis identifies the misspelled “athiest” as a thing rather than a person “-ist”. Since people are not things, he refuses to believe it.
17. “Which telephone?” Arckon thinks “the” indicates a specific telephone, but doesn’t find one mentioned earlier in the conversation.
18: “What does a b stand for?” This is really just an “error message” that may sound intelligent but totally isn’t.

 

 
  [ # 23 ]

Well, I didn’t expect Izar to do that well either!  I’m rather shocked - this bot is only 17 months old!  However, I am extremely excited with the direction of questions this year.  I’ve long been an advocate of changing how bots were judged.  The Loebner Prize is awarded to “the chatterbot considered by the judges to be the most human-like.”  In my humble opinion, I don’t think a “Trivia-Bot” or a “Math-Bot” satisfies that definition.  I know tons of many special humans that can’t answer literally any typically asked math or trivia question, yet they are more “human-like” than any bot out there.  Also, if we were to stumble across a new tribe in the jungle, take the time to learn their language, and ask them questions that are typically used in other contests they would not be able to correctly answer 95% of them.  Does that mean they are not “human-like?” Not at all.  I could also ask an 7-year old child the typical questions asked during other contests, and once again they would fail to answer correctly the majority of them…yet they are more “human-like” than any other bot out there. I could go on giving examples, but you get my point.

I developed Izar (the brains behind the Android “Back Talk” application) to hold intelligent and fun conversations. His persona is an alien that has recently arrived to our planet to study us.  The Alien does know some things that he has studied, but only as much as you would expect.  Izar fails more often than not in word game questions as would most “aliens” that have just arrived.  I know there are better “bots” out there than Izar, but I didn’t develop Izar as a “bot,” rather an animated talking friend. I think Steve would agree that Mitsuku also falls under this definition.  I believe that’s what separated these bots out a bit from the pack and allowed them to do as well as they did.

I want to send out a big congrats out to Mitsuku, Tutor, & Rose! You undoubtedly have spent a thousand or more hours and years than I have developing your technologies. I am humbled by the temporary top spot, but know that I know your achievements far outshine my own - this round of questions happened to fall slightly my way.  I have a LOT of work to do between now and September!

 

 < 1 2
2 of 2
 
  login or register to react