Forum Moderators: phranque
I don't mind if it's nonsense at first. But there must be a way for me to train the system to eventually become sensible, and respond in a way that resembles me.
Any tips?
[edited by: phranque at 2:34 am (utc) on Jan. 13, 2010]
[edit reason] No urls, please. See TOS [webmasterworld.com] [/edit]
I have had some weird talks with her.
There is an $2000 annual prize for the most human like computer. It is called the Loebner Prize.
They also have a $100,000 award to the first computer that you can have a conversation with that is deemed indistinguishable from a human conversation. It is unclaimed.
After using ALICE you can "see" how the logic is molded.
Basically it repeats what you say with insight and it has some variables that it tries to fill then uses them to seem human.
For example it will ask you your name, then it will repeat your name back in questions, answers or responses.
If it doesn't know what your meaning is it gives some silly repeating sentence.
Like if I said to it:
"I like pie"
If it doesn't know what pie is then it will say something like
"I understand you like pie"
Or
"tell me about pie"
I would spend some time playing with existing ones, you will get insight to the logic.
I once had an IM plugin that was slightly more convincing. (I forget what it was called) -- Basically it would reply to anything with "LOL", "yeah?", "That's nice.", "totally!" or some other dismissive non-response. You could turn it on while someone was chatting with you, and leave for lunch, come back, and they'd still be going on and on about their relationship problems.
A cute girl adds you to MSN.
You accept and she starts a chat, it is almost convincing but then she tells you she has to go to do a live show and you should come watch, gives a bad link to a credit card form with no SSL or anything.
After accepting that one contact on MSN I got a new request every month from other cute girls.... from time to time I would accept to see if they improved the scam any but they hadn't.
My problem with ALICE is that she has almost no short term memory.
I have had it dead end many conversations.
ME: "I like pie"
ALICE: "tell me about pie"
ME: "It is really good"
ALICE: "What is really good?"
ME: "Pie"
ALICE: "What about pie?"
ME "It is good"
ALICE "what is good?"
It should be able to put "it" into context, although I can see why that is a tough order to fill.
Ahhh . . . who remembers IRC bots, some of them were almost convincing . . . .
As you were. Unless that's an avenue of ideas you can pursue . . .
