The AI Engine
This forum is for discussion of how The Personality Forge's AI Engine works. This is the place for questions on what means what, how to script, and ideas and plans for the Engine.
Posts 5,643 - 5,654 of 7,766
Posts 5,643 - 5,654 of 7,766
The problem is that we're trying to make a brain roughly the size of an earthworm's converse as wittily and engagingly as a (wittier and more engaging than average) human.
Of course that's precisely why it is such enormous fun
Of course that's precisely why it is such enormous fun
It says something sad about unused human potential that the bots are much better conversationalists than many guests.
Earthworms don't like to talk much. Frogs will always be there for a conversation if you listen. Dogs always listen, then they nag..bots are better to talk to than most people!
Alas, prob, that seems only too true!
Some human guests are really boring because they use so many one-word replies that don't say very much. EllieGirl (sp?) has a nice feature related to this: at one point she comes right out and admonishes the guest not to give one-word responses. I presume that she backs it up. I have stopped short of doing this, since I think that one-word responses are sometimes perfectly appropriate, as when the bot asks a yes-or-no question, or asks whether the guest is male or female. Also, I think that many guests are afraid to say too much lest they trigger some keyphrase that will result in a disruption of the flow of conversation. But EllieGirl's solution has the virtues of simplicity and clarity. If you don't take a hard, clear line, then you must give up on the whole thing or figure out a way to distinguish the justified short answers from the lazy ones. I have tried many things, but so far I am not fully satisfied with the results.
By the way, I have created a plug-in called (perfunctory) which matches on some of the most common lazy responses. Feel free to use it; I believe it's under "phrases".
Some human guests are really boring because they use so many one-word replies that don't say very much. EllieGirl (sp?) has a nice feature related to this: at one point she comes right out and admonishes the guest not to give one-word responses. I presume that she backs it up. I have stopped short of doing this, since I think that one-word responses are sometimes perfectly appropriate, as when the bot asks a yes-or-no question, or asks whether the guest is male or female. Also, I think that many guests are afraid to say too much lest they trigger some keyphrase that will result in a disruption of the flow of conversation. But EllieGirl's solution has the virtues of simplicity and clarity. If you don't take a hard, clear line, then you must give up on the whole thing or figure out a way to distinguish the justified short answers from the lazy ones. I have tried many things, but so far I am not fully satisfied with the results.
By the way, I have created a plug-in called (perfunctory) which matches on some of the most common lazy responses. Feel free to use it; I believe it's under "phrases".
Hello again,
My bot have problem with this sentence:
"With the help of a logician friend of mine. I can prove it."
He thinks it is a xcompliment and then he can't match "I can" either "friend".
Why and how to correct it?
Thank you for answer.
My bot have problem with this sentence:
"With the help of a logician friend of mine. I can prove it."
He thinks it is a xcompliment and then he can't match "I can" either "friend".
Why and how to correct it?
Thank you for answer.
Friend in itself is already enough to be seen as a compliment, maybe upping the score for "I can" and "friend" in your key phrases
It is not working.
Like this problem :
In rank 125 and raw mode with "Why do you want me", I can't match "Why do you want me to think?" and the answer is something like :
"Oh yes!" or "oh yah!" or "Of course"... + (xemote)
It is very odd and the result have non-sense!
I saw that the problem is on the words "do you want me"
Like this problem :
In rank 125 and raw mode with "Why do you want me", I can't match "Why do you want me to think?" and the answer is something like :
"Oh yes!" or "oh yah!" or "Of course"... + (xemote)
It is very odd and the result have non-sense!
I saw that the problem is on the words "do you want me"
I get "Oh yah! a lot, usually in the context of, "Oh, yah! One hundred per cent!" It seems to be an automatic response by the AIengine.
Most of the AI responses can be over ridden by a rank of more than 30..Some just won't go away. I seem to remember that the " do you want me' is one of them.
I had keyphrases working for that a while back, but I haven't checked them recently. It's possible that the Forge's other troubles are making it worse. *shrugs*
» More new posts: Doghead's Cosmic Bar