The 10 questions from the CBC have been asked and while I don’t think I’m allowed to post the actual questions yet, I was wondering how your bots found them?

I personally was pretty pleased by the way Mitsuku answered, the only annoyance being about the question regarding the age. Had the judge just used the year instead of the full date, Mitsuku would have answered correctly. Ah well

Good luck to all.

My biggest disappointment was that Morti is, based on those 10 questions, still 40% ALICE. Well, was, at least. He’s already been “fixed”.

I noticed, too, that one question did NOT get answered in a manner that it was SUPPOSED to, due to a category still being in place that I thought that I had removed. Oh, well. Overall, Morti didn’t do as well as I had hoped, but I feel that he’s possibly still in the top 50% of the scoring. I’ll have to wait and see, though.

(Morti failed the age one, too, Steve)

I’ve amended the age one so Mitsuku can handle it now. I will amend the bornin.aiml file on my AIML page with the changes so it can handle the full date now instead of just the year. Mitsuku also failed on the shapes one too. I had coded up to hexagon but no further. Lazy me but I’ve now fixed it (the AIML, not my laziness!)

I’m hoping they score extra for additional functions. When the judge what a certain website, Mitsuku explained what it was and opened the site for him. I hope this gets an extra point!

My only other problem was with the last question about Jeopardy. Had they asked “Who is Watson”, she would have replied correctly.

I’m now looking forward to reading the chatlogs from the other bots and hopefully preparing for round 2

There’s a Round 2???!?!?!?!?!?

Aren’t we getting a bit close to the “no discussing the inputs/responses” rule? Or are we in the clear already? I’d hate to have Mitsuku or Morti banned for life. I’m even afraid to discuss Morti’s “new responses” right now, so I’ll hold off for just a bit.

Yes, we had probably not discuss any more until we get the all clear from Wendell. I’m assuming this first round of questions was to get the final 10 bots to judge? Eliminating the ones who only talk about Pi etc.

Well, hopefully, if this first round was to separate “the wheat from the chaff”, that Morti will survive to participate further. I had the impression, though, that more than one judge would be asking the same questions. Apparently I was incorrect in my assumption. Guess I’ll have to carefully re-read the rules.

Steve Worswick - Mar 22, 2011:

I will amend the bornin.aiml file on my AIML page with the changes so it can handle the full date now instead of just the year.

You’re a good man, Steve! please holler when you get it updated. I re-downloaded it earlier this afternoon, but it’s not the “new and improved” version yet.

I think aici did very bad, most of the sentence structures were not yet known. Ah well it was a gamble…

I’ll be interested in comparing our respective rankings, Jan. Not for any sort of “competitive” reasons, mind. Our bots are on different development paths, after all. but still, I think it will be interesting to see how we each did.

By the way, this is how I would have “scored” Morti’s responses, were I a judge, and not involved with him:

Dave Morton - Mar 22, 2011:

I’ll be interested in comparing our respective rankings, Jan. Not for any sort of “competitive” reasons, mind. Our bots are on different development paths, after all. but still, I think it will be interesting to see how we each did.

They used the client version, so I’m not 100% certain. I do know some judges had problems with it, and it got stuck on at least 1 question, so things were dismal.

You weren’t one of the two bots that “had technical issues”, were you?

Dave Morton - Mar 22, 2011:

You weren’t one of the two bots that “had technical issues”, were you?

yep.
1 judge didn’t seem to have enough permissions on his machine to install the .net libs, so he was out. Another one managed to start the client without the network loaded : the client loads a network like another app loads a file from the prompt: as a command argument, so something must have gone wrong there, either the path to the network in the argument was incorrect, or the app got started without arguments, anyway, he was out as well. Take 3 seemed to have worked better, but even the 1st question was, from a structural point of view, very difficult. Perhaps question 8 and 9 would have worked, all the rest were ouside of Aici’s curent scope: 2 consecutive nouns weren’t working yet, ‘NP, NP or/and NP’ wasn’t working yet, ‘Main clause-who- sub clause’ isn’t tested, don’t know if it works, date input isn’t implemented yet, haven’t created any frames yet for ‘there’,  and then there are only 2 questions left.

Well, as they say, if it were easy, it wouldn’t be as much fun!

That said, I know how it feels to have “too much fun”.  I’m truly sorry that AICI had difficulties. With any luck, he/she/it will have scored better than “the PI bot”.

If one of the questions had been “what is pi”, I would have fallen off my chair!:lol:

Dave Morton - Mar 22, 2011:

I had the impression, though, that more than one judge would be asking the same questions. Apparently I was incorrect in my assumption.

I think you were correct. Part of the rules say:

Judging
We will select judges from the general public. Between 3/15/11 and 4/01/11 all the judges will speak with all the bots entered. The judges will work independently of each other. The bots will be asked a sets of 10 questions. Each question will be spelled and phrased exactly the same way. This is to insure that each bot gets a similar conversation. The questions will be asked within the framework of a conversation. The judges will steer the conversation via the questions asked but will also follow the leadof the bot.

However, I guess with 40+ bots, this wasn’t very practical, especially as many of them needed to be downloaded and installed. I tried talking with all of the bots but was unwilling to download things like itunes and AOL messenger to talk to some of them.