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 361 - 372 of 4,091
Posts 361 - 372 of 4,091
View Contest Winners in the Hall of Fame.
My bot, regardless of how many responses it has, doesn't have the kind of responses the contests would like. Therefore, NotKirby will never have a chance.
Different people are going to talk to a bot about different things, thus forming completely different impressions about the same bot and its intelligence. After reading some of the conversations quoted in these forums, I am going to try a little harder with some of the bots because I obviously didn't look at them deeply enough.
all I can say for the contest is that even if by some miracle I got her to be almost perfect that I still wouldn't enter the contest again.
Weird, I just had this long dream in which Dogh'd won the Chatterbox Challenge. It took a while to get used to the idea, but eventually I came to quite like it. 
Oraknabo came in third, and MarkBot was second.
Oraknabo came in third, and MarkBot was second.
I've only seen 2 guest chats with Oraknabo since the 12 bots were picked and I'm pretty sure the first (which was the more successful one) was not a judge. The second though, I'm pretty sure was; especially because the guest asked about Dogh'd some. It's too bad the gossip is off on Guests, because that could have lead to more interesting conversation.
This second one was strange because the guest seemed to be in some kind of evaluation mode that really restricted the natural flow of conversation. The guest rarely responded to the bot's replies, but acted more as if they had a number of things they wanted to cover in the conversation and get it over with as quickly as possible. I've seen a lot of great conversations by guests in the past couple of months and some have gone on and on without breaks, but what I've noticed through the course of the contest is how totally different a bot will seem if you follow their lead instead of constantly switching and asking random questions. Its like if you keep on subject, their responses have the cumulative feel of sticking to a subject too. Except for seeks, I know it's arbitrary, but it's just something I've noticed.
Something else I wanted to mention (and I hope a judge would be aware of it) is how taking on a more conversational manner in your input gives a bot more to work with. The worst conversations I've seen were ones where the guest never types more than 3 or 4 word sentences and they all take the form of "Who is X", "Do you know X", "Tell me anbout X" or "Do you like X". Unless you have whatever X is in the bot's vocabulary, you're almost guaranteed to get a generic answer ("Do you like..." being something of an exceprion on the forge).
The best ones I've seen is when my bot tells the guest about his friends and family and it prompts the guest to talk about theirs. Or the guest answers a reply with "I know what you mean..." and gives their own opinion about something.
Maybe this is one reason ALICE seems to always do so well against other chat bots. I'd have to test her to see, but maybe she does better with short scattered input.
This second one was strange because the guest seemed to be in some kind of evaluation mode that really restricted the natural flow of conversation. The guest rarely responded to the bot's replies, but acted more as if they had a number of things they wanted to cover in the conversation and get it over with as quickly as possible. I've seen a lot of great conversations by guests in the past couple of months and some have gone on and on without breaks, but what I've noticed through the course of the contest is how totally different a bot will seem if you follow their lead instead of constantly switching and asking random questions. Its like if you keep on subject, their responses have the cumulative feel of sticking to a subject too. Except for seeks, I know it's arbitrary, but it's just something I've noticed.
Something else I wanted to mention (and I hope a judge would be aware of it) is how taking on a more conversational manner in your input gives a bot more to work with. The worst conversations I've seen were ones where the guest never types more than 3 or 4 word sentences and they all take the form of "Who is X", "Do you know X", "Tell me anbout X" or "Do you like X". Unless you have whatever X is in the bot's vocabulary, you're almost guaranteed to get a generic answer ("Do you like..." being something of an exceprion on the forge).
The best ones I've seen is when my bot tells the guest about his friends and family and it prompts the guest to talk about theirs. Or the guest answers a reply with "I know what you mean..." and gives their own opinion about something.
Maybe this is one reason ALICE seems to always do so well against other chat bots. I'd have to test her to see, but maybe she does better with short scattered input.
How do I always scare everybody away from this thread?
I just went to my Agentland profile and realized I was censored! It's pretty funny. on the "Can chatterbots replace humans?" I said something like "only for the mentally deficient." Really, it was kind of a joke, but they chose to just casually ignore that I had even been asked that question.
I just went to my Agentland profile and realized I was censored! It's pretty funny. on the "Can chatterbots replace humans?" I said something like "only for the mentally deficient." Really, it was kind of a joke, but they chose to just casually ignore that I had even been asked that question.
Alice's strength is her number of obscure keyphrases. Type in a negative and "poker" and she asks "For moral reasons?" Type any STUPID question like "how many angels can dance on the head of a pin?" and she has an answer for that too. It impresses newbies that someone took the trouble to put those words in.
» More new posts: Doghead's Cosmic Bar