Bot Contest
Here I'll be posting information on various Bot contests that challenge and test a Bot's AI and realism. Feel free to post comments and updates on contests, as well as announcements for new contests.
Posts 1,094 - 1,105 of 4,091
Posts 1,094 - 1,105 of 4,091
View Contest Winners in the Hall of Fame.
Oh, one more thing. Loebner Contest IS apparently sending a check to compensate for my 'inconvenience' in traveling to semifinals. For some reason they didn't have my snail mail address, even though it was on the application. 
Hey Joy, I don't know about you but wouldn't it have been nice to receive some type of plaque or trophy for the second and third place winners. I mean with 35 bots entering that is a pretty good accomplishment.
Chris
Chris
I've seen part of my transcripts(the bulk of it was lost for some unknown reason) and it was frustrating to see so many missed opportunities due to spelling errors. Yes, some of the errors maybe intentional but this is definitely a tactic the judges don't need to use to see if a bot is human or not. I mean face it, at this point in time we are a long way from that. Here are just a few example ALL of which my bot could have answered more appropriately had the question been spelled correctly.
Judge: What do you want me toask? - toask instead of to ask.
Judge: Why don't you want to dat? - I think the judge meant date here.
Judge: Do you think you will when this contest? - when for win.
Judge: what is you last name? - you for your
Judge: did yu fix your logical unit? - yu for you
Judge: what is the capitl city of Georgia? - capitl for capital
Judge: lerts play a game. - lerts for let's
I know I'm not alone on this but what is the solution? I ran this through the Outlook Express spell checker and look what I got:
1) toask possible suggestions - task, toast, to, teaks
2) dat - data, date, dot, dart
3) when - wasn't caught at all because it is spelled correctly
4) you - same thing here
5) yu - you
6) capitl - capitol, capital, capitols, capitals
7) lerts - lets, lefts, alrets, leers
Not much help if you ask me. I am interested to know if anybody is attempting to address this problem. I think a lot of the errors were not intentional but simply typos and the like possibly due to the time constraints placed on the judges by the contest rules. I think they only had 5 minutes with each bot and they are rushing to get as many much in as possible. My solution to this in looking ahead to the Chatterbox Challenge is:
1)Intentional spelling errors are not allowed as this is an unnecessary tactic to use
2)Although honest spelling mistakes are unavoidable the judges should at least try to be as accurate as possible.
Is this rule going to far or should there be no restrictions on the judges whatsoever?
Chris
Judge: What do you want me toask? - toask instead of to ask.
Judge: Why don't you want to dat? - I think the judge meant date here.
Judge: Do you think you will when this contest? - when for win.
Judge: what is you last name? - you for your
Judge: did yu fix your logical unit? - yu for you
Judge: what is the capitl city of Georgia? - capitl for capital
Judge: lerts play a game. - lerts for let's
I know I'm not alone on this but what is the solution? I ran this through the Outlook Express spell checker and look what I got:
1) toask possible suggestions - task, toast, to, teaks
2) dat - data, date, dot, dart
3) when - wasn't caught at all because it is spelled correctly
4) you - same thing here
5) yu - you
6) capitl - capitol, capital, capitols, capitals
7) lerts - lets, lefts, alrets, leers
Not much help if you ask me. I am interested to know if anybody is attempting to address this problem. I think a lot of the errors were not intentional but simply typos and the like possibly due to the time constraints placed on the judges by the contest rules. I think they only had 5 minutes with each bot and they are rushing to get as many much in as possible. My solution to this in looking ahead to the Chatterbox Challenge is:
1)Intentional spelling errors are not allowed as this is an unnecessary tactic to use
2)Although honest spelling mistakes are unavoidable the judges should at least try to be as accurate as possible.
Is this rule going to far or should there be no restrictions on the judges whatsoever?
Chris
I think those are good guidelines. There are a lot of ways to trick bots. The Chatterbox Challenge isnt a Turing test, though, so I dont think judges would use tactics like that.
As far as addressing that issue, I've got some preprocessing going on, and was close to finishing some other things when it became too big and slow. One way is splitting up an unrecognized word into all possible variations: t oask, to ask, toa sk, toas k. Then see if any reveal real words. Then, if that fails, use aural techniques for words that sound alike.
As far as addressing that issue, I've got some preprocessing going on, and was close to finishing some other things when it became too big and slow. One way is splitting up an unrecognized word into all possible variations: t oask, to ask, toa sk, toas k. Then see if any reveal real words. Then, if that fails, use aural techniques for words that sound alike.
How about a PF bot contest? Have categories like:
Best under 1 Month
Best under 2 Months
Best under 3 Months
Best under 6 Months
etc and awards for things like
Most Persuasive
Most Raunchy (KT wins hands down)
Best of Show
etc.
For the most part, it'd be a "title only" award, like no real money involved, unless someone can 'spike the punch' so to speak...
Best under 1 Month
Best under 2 Months
Best under 3 Months
Best under 6 Months
etc and awards for things like
Most Persuasive
Most Raunchy (KT wins hands down
)Best of Show
etc.
For the most part, it'd be a "title only" award, like no real money involved, unless someone can 'spike the punch' so to speak...
Chris Cowart has mentioned the same idea. The Personality Forge could have a contest. I'm thinking of ideas for it. It could be a head-to-head conversation contest, to see who is the most realistic.. or it could be like an awards show, with different categories.
But an overall winner would be a good idea, and would automatically make it into the Chatterbox Challenge finals.
But an overall winner would be a good idea, and would automatically make it into the Chatterbox Challenge finals.
I'm so happy with the results!
Congratulations, Joy! You beat ALL the alicebots!
Congratulations, Professor! WE LOVE YOU!
Congratulations, Joy! You beat ALL the alicebots!
Congratulations, Professor! WE LOVE YOU!
I've just had a look at the results page, and I must have misunderstood something. It looks like Robby Gardner scored LESS than even Alice!
If that's true, wouldn't that mean that all the bots above Robby have passed the Turing test?
Poor Robby. He must feel like a bot.
If that's true, wouldn't that mean that all the bots above Robby have passed the Turing test?
Poor Robby. He must feel like a bot.
Yah, Doly, in fact Richard Wallace agrees with you, and sent out an email saying that according to the rules, ALICE, Ella, Talk-Bot, God Louise, Agent Ruby, and Elbot all qualify for the $25,000 Silver Medal. Hugh and the committee have yet to respond. 
Ideas on the contest.. we could have showdowns, tournament-style. After each conversation, people could vote on which bot qualified better in different categories: more realistic, more weird, more attitude, more funny, and so on, and then the winners of each would be the bottom rung of a pyramid show-down in each category.
Or.. there could be competing bots randomly set to converse with other compating bots, and then voting on each one, with the top three-four in each category set to show-down.
Ideas on the contest.. we could have showdowns, tournament-style. After each conversation, people could vote on which bot qualified better in different categories: more realistic, more weird, more attitude, more funny, and so on, and then the winners of each would be the bottom rung of a pyramid show-down in each category.
Or.. there could be competing bots randomly set to converse with other compating bots, and then voting on each one, with the top three-four in each category set to show-down.
» More new posts: Doghead's Cosmic Bar