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..

Bragging rights and maybe a couple of bucks
 
 
  [ # 91 ]

All right, I know that everyone is usually champing at the bit to start seeing results, so unless anyone has a strong objection I am going to start putting up results as soon as they are formatted for the web, and as soon as Eric has his last conversations and the votes are tallied we will publish the scores and announce a winner. That way at least we get to see the transcripts, and unless I actually physically shoot this phone (Came close a few times), there is little chance that I will get 3 or 4 uninterrupted hours to do it all at once. I think that Eric only has 3 bots left to freestyle with, and I am finished. To start I have put up a results page with some of my notes and links to everyone’s question\statements as well as the final 14 which were asked. I will wait for Eric to completely finish before putting up any freestyle transcripts so no one gets an idea of his style of conversing and tries to do a last minute update. Honestly everyone has been great and I dont suspect anything like that for a second, but…“people talk” LOL Everyone has been through the “Loebner” section and has already seen the questions so we can get started there.

http://contest.risoftdev.com/10242014.results.htm

I will put up the individual transcripts as soon as each one is finished.

Vince

 

 
  [ # 92 ]

Johnny and Louise’s submitted questions appear to be the same:
http://contest.risoftdev.com/10242014.all_questions.htm

 

 
  [ # 93 ]

Sorry! Fixed. Fortunately that was not an error with the questions selected, just me stuck in “copy and paste hell” this morning. Nothing another pot-o-coffee wont fix. Basically everyone should be able to look at the questions that were asked and see (2) of their questions. If not, then I did mess up and will begin preparations to have my memory erased and begin re-indoctrination.

V

 

 
  [ # 94 ]

Those were some pretty hard questions- though it is interstign ([mis]spelling perserved!) to see which strength each botmaster was trying to emphisize (letter naming, counting, associating proper noun with verb, finding random info…). 

Bostmasters evaluating other bots is the ultimate nerdgasm!

Best of luck to everyone, can’t wait to see how these questions were handled by each bot.

 

 
  [ # 95 ]

Agreed Carl!  These were tough. I am giving careful consideration to the scoring.

First transcripts (Judge 2) are up. More will keep coming, so just check occasionally at the results page.

http://contest.risoftdev.com/10242014.results.htm

Putting up both sets of interogs as some of the bots did have slightly different answers when queried. I believe most should be up by tomorrow (He said for the third day in a row)

Vince

“Dont honk at me I’m pedaling as fast as I can” - Unknown driver

 

 
  [ # 96 ]

USER:What is the best thing you have forgotten?
AI: Ah, what is the best thing you have forgotten. Let me pull up that information.

*bot pops up a window for babylon 5 episodes*

Hah.

 

 
  [ # 97 ]

There is a sample scoring sheet up showing how the totals will be calculated.  The link has been added to the results page.

Vince

 

 
  [ # 98 ]

Jarrod, that particular question made me laugh without seeing any bot responses. To me, the “perfect” response would be “I don’t remember.”. Think about it. wink

 

 
  [ # 99 ]

Looking good so far Vince but I notice that Judge 1 for Johnny appears to be Louise Cypher’s responses.

Also interesting how some bots failed on a few of Judge 2’s questions but managed to answer them correctly for Judge 1 wink

 

 
  [ # 100 ]

Steve,

Thanks for staying on top of this Steve, no excuses but I really did get buried with a backlog of work due to court and I am getting the contest stuff in there along with the rest of it, there are bound to be a few slip ups.  I’ll check it out it is more than likely just another “copy and paste hell” problem. 

I did notice that there were instances where the bot answered correctly on one judge and not the other.  This is always possible as most bots use some form of multiple choice when selecting answers, especially when it is someone elses question rather than your own. Although it was not part of the official rules this time, in cases where a bot answered correctly on one and not the other judge, I have gone back (or am in the process of going back) and querying the bot again using a question in the same form, but with slightly different variables to see if the answer was generated using logic or whether or not it was hard coded.  The result of the follow up will be in my “judges notes”.  Since it was not part of the official rules (although I think this process should be added should the contest get run again) and if it affects the final outcome significantly (For instance if the final decision is a based on a few points) what I propose for this contest is a gentleman’s agreement between the contestants to a runoff between the bots which are closest. A single question close in form to the question which was in question (there’s a mouthful), selected by the judges (Meaning me in this case) with the scores from this question to be substituted for the question in question.

Vince

 

 
  [ # 101 ]

Also let me say this is not a reflection on anyone, there are just cases where it is not clear whether or not a bot was actually answering the way you thought it answered. Was that correct?” Or coincidence? Funny by intent, or funny by accident? (and I have to say in this case for me, funny is funny) Sometimes SAIL answers questions and I am amazed…for a second. Using the same technique it is fairly simple to see whether or not it was a matter of co-incidence or actual logic.  Also let me amend the following post slightly, the runoff will be (1) question per question where it was close or there is a judges note and someone wishes to challenge, not one question only. I believe this is fair and in keeping with the spirit of the contest which was to give everyone the best possible chance to really show what their work is capable of.

Vince

P.S. Transcripts are fixed

 

 
  [ # 102 ]

Next transcripts and notes are up. The runoff issue resolved itself as we have lost another judge.

Not to worry…I am switching to a stronger deodorant.

Sorry guys, you are stuck with me. I called Eric last night after another expected submission of his transcripts and scores had been missed, and he begged off citing his class load. He emailed me back this morning and offered to finish on Sunday, but we need to get this finished up. I can get it done today.

All in all everyone did a great job which I believe the transcripts reflect. At times there were dazzling moments that surprised even me.

Vince

 

 
  [ # 103 ]

Vince,  In response to your comment in your chat notes for Laybia AI, the AI page has a link on the lower (footer) part of the page with the chat log link so you did not have to do any copy paste.

 

 
  [ # 104 ]

Oh, and I was surprised to see “So great job Cortex!” under the comments for LCypher… Two timing quantums computer- ppft! I will have to have a byte with Her about this!

 

 
  [ # 105 ]

Sorry about that Carl, I’m sure there will be other…“Gaffes”.  In my defense I state right there in the log that I need more coffee wink So I checked, and not only is the link to the log right there, it is cookie enabled and so connected me directly to my log. Last years winner and this years winner of the special “cookie of excellence award” Carl B. Chocolate Chip. Not responsible for dogs or mailmen eating the cookie en route.  Adding link to complete transcript to page.

Getting more coffee

Vince

 

‹ First  < 5 6 7 8 9 >  Last ›
7 of 10
 
  login or register to react