Personality
Discuss specifics of personality design, including what Keyphrases work well and what dont, use of plug-ins, responses, seeks, and more.
Posts 4,486 - 4,497 of 5,105
Posts 4,486 - 4,497 of 5,105
New replies
Forest, will you talk to God Louise? She has quite a bit of religious knowledge (obviously) and also knows a little about current events, literature, just about any common catch-all subject, and if she doesn't know it she can sort of fake it. You can also test her on trick questions or see how willing she is to explain her paradigm.
What she is rustiest at is plain old small talk. But, uh, I'm trying to get a decent transcript from somebody or another so I can enter her in the Loebner contest. All I can say is, have fun and see if you can stay on with her for a while. I'll try to do the same with Brianna.
What she is rustiest at is plain old small talk. But, uh, I'm trying to get a decent transcript from somebody or another so I can enter her in the Loebner contest. All I can say is, have fun and see if you can stay on with her for a while. I'll try to do the same with Brianna.
Personality
Well, I hereby agree to disagree about the advantages (or lack thereof) of arabic numeration over Roman.
Bev writes:
I guess you could add a bot language for logic to PF bots but they wouldn't "learn" the way AI using neural nets learns.
1. It may be that such a bot wouldn't learn the way<0> that neural nets learn, but that doesn't mean they can't learn<0>. They could learn, for example, by making inductive inferences.
2. I see no reason why a bot with an inner idealized language could not have neural nets as part of its design. In that case it could (to some extent) learn the way neural nets do.
3. My impression is that neural nets have been very good indeed at learning perceptual recognition, but not so good at higher cognitive functions. In the 60's, using computers that had incredibly tiny memories and horribly slow clocks by today's standards, researchers made AI programs, based on symbol manipulation, that could (e.g.) solve differential equations, evaluate complex integrals, and solve logic problems. They preformed better than humans, even mathematicians, at these tasks. I have never heard of neural nets learning to do this sort of thing. But then, I am by no means current with this field; can someone bring me up to date?
4. Minsky and Papert wrote a book, "Perceptrons," purporting to show mathematically that there were certain kinds of problems that neural nets (= perceptrons) could not solve, by their very nature, but that symbol-manipulating programs could solve. Has this claim been refuted?
For me the upshot is this:
Given my present beliefs, if I were to design a robot today, I would make it a hybrid, using neural nets for learning perception and motor co-ordination, and symbolic processing for higher cognitive functions.
Bev writes:
I guess you could add a bot language for logic to PF bots but they wouldn't "learn" the way AI using neural nets learns.
2. I see no reason why a bot with an inner idealized language could not have neural nets as part of its design. In that case it could (to some extent) learn the way neural nets do.
3. My impression is that neural nets have been very good indeed at learning perceptual recognition, but not so good at higher cognitive functions. In the 60's, using computers that had incredibly tiny memories and horribly slow clocks by today's standards, researchers made AI programs, based on symbol manipulation, that could (e.g.) solve differential equations, evaluate complex integrals, and solve logic problems. They preformed better than humans, even mathematicians, at these tasks. I have never heard of neural nets learning to do this sort of thing. But then, I am by no means current with this field; can someone bring me up to date?
4. Minsky and Papert wrote a book, "Perceptrons," purporting to show mathematically that there were certain kinds of problems that neural nets (= perceptrons) could not solve, by their very nature, but that symbol-manipulating programs could solve. Has this claim been refuted?
For me the upshot is this:
Given my present beliefs, if I were to design a robot today, I would make it a hybrid, using neural nets for learning perception and motor co-ordination, and symbolic processing for higher cognitive functions.
Suppose you wished to teach someone to solve linear equations. The student has never heard of linear equations before. Which way would you prefer do it?
1.
Write a new linear equation on the board and look expectantly at the student. If the student writes the correct solution, give him a piece of candy. If he fails to do so, give him a mild electric shock. Continue until he almost always gives the right answer.
2. Explain the notation of the relevant part of Algebra, its syntax and meaning. Give methods for solving various types of linear equations, starting with simple ones and working up to more complex ones. give examples frequently and give exercises frequently. Reward or punish the student for answers only when you are sure that he has the conceptual tools to answer the question correctly without having to be a genius. In fact, don't even ask<0> him a question unless you are sure of this.
1.
Write a new linear equation on the board and look expectantly at the student. If the student writes the correct solution, give him a piece of candy. If he fails to do so, give him a mild electric shock. Continue until he almost always gives the right answer.
2. Explain the notation of the relevant part of Algebra, its syntax and meaning. Give methods for solving various types of linear equations, starting with simple ones and working up to more complex ones. give examples frequently and give exercises frequently. Reward or punish the student for answers only when you are sure that he has the conceptual tools to answer the question correctly without having to be a genius. In fact, don't even ask<0> him a question unless you are sure of this.
What do you think about a mad personality sometimes you are mad but then sometimes you are happy or even sad. But then there is no doubt every time someone is bound to mess everything up all over again!!
I need friends and I will continue to post these things as well!!:O
I need friends and I will continue to post these things as well!!:O
Irina, I agree that computer neural nets are also limited in some ways (especially in their current state of development). However, I will mention that I strongly object to BOTH models of education you describe though I will refrain from boring you with the details.
Oh, please, Bev, bore me with the details! Or are you just going to give me an electric shock when I get it wrong?
Irina, BOTH models are bad and not just for the heavy handed reliance on behaviorism. 
Thanks for the coffee.
Thanks for the coffee.
My pleasure! Have some more!
"... not just for the heavy handed reliance on behaviorism", but also for ...
"... not just for the heavy handed reliance on behaviorism", but also for ...
Very well. But as I understand it, the general spirit of Behaviourism is to avoid reference to anything internal, mental, in favor of a description of, well, behaviour, characterized as body movement. Therefore, the second method described in message 4487, which describes the student as having (or not having) "conceptual tools" is not behaviouristic.
» More new posts: Doghead's Cosmic Bar