Ai Forums Home Welcome Guest    Saturday, November 25, 2017
Ai Site > Ai Forums > Hal and other child machines > Question about iHal/Hal feature Last PostsLoginRegisterWhy Register
Topic: Question about iHal/Hal feature

AXR
posted 1/5/2011  07:05Send e-mail to userReply with quote
Hi, I was just wondering:once the bot is past its basic training, which I realize takes a while, is there any way to teach a Hal to respond to a specific series of messages? For example:

Bot> Would you like me to suggest a topic?
User> sure

And then the bot suggests a topic. I am not planning on using it in this scenario but I'm sure I'll come across something more specific, that I would like it to give a specific response for, depending on what the user says. Please let me know if I need to explain it differently.

Also, (another question), is developing a bots personality basically just based off of the input you give*, or is there something else to develop it as well.

*For example, if I wanted my bot to be witty, I would train him using witty responses.

Thanks,
AXR


yaki
posted 1/7/2011  16:48Send e-mail to userReply with quote
You know what they say: Garbage in, garbage out! (GIGO). If you teach your bot witty responses, you will have a witty bot. Wit does not come from thin air. Wit is a skill, a delicate, subtle skill of using language.

As for your bot making independent suggestions (of discussion topics, for example). There is no "randomness". a "random" event (such as a random topic suggested by the bot), is always brought about by causes unknown (to the person who considers it random).


AXR
posted 1/7/2011  19:27Send e-mail to userReply with quote
Thanks, I just wasn't sure, because on the Virtual Personalities>Hal page it listed personality and responses separately. I'm actually not trying to train a witty bot, but rather a bot who thinks he's in the military and talks like a humorous superior. It sure sounds like it will be hard to built such a bot by starting from the basics of language!

:) I realize that many things that bots say are not random. However, I was wondering if there was any way to program the bot so that he gave a random response out of those that were taught to him. For example, giving the bot more than one response to "hello" and then it would select one out of those.

Also, what I was having trouble explaining in the first post, was what I was saying about patterns. It was about my bot giving an answer based on what the user entered, AND what he (my bot) had said just prior to that.

For example, my bot's default response to the input "yes" might be "yes?" But say that before that my bot had asked a question such as "Do you like the color red?" If the user said "yes", could I have the bot give the response "Cool, I like red too," by some sort of code? Or would more advanced responses come with time as my bot began to understand the language?

To further clarify, I cannot just enter the response to "yes" to be "Cool, I like the color red too." If I did this, of course the bot would likely use that phrase out of place many times. So, I was wondering if there is some sort of conditional statement I could use such as the following, such as:
if(lastbotstatement=="do you like red?" & lastuserstatement== "yes")

then(saythisphrase)

I realize I cannot use a programming language with the bot, but I wondered if there is any built in method to achieve a conditional response, such as the one above.

Thanks a ton!
AXR


Spydre
posted 1/11/2011  07:54Send e-mail to userReply with quote
In response to your question about your bot learning "context" based answers (in reference to "yes?" or an answer about the color red instead), emphatically yes. HAL will learn contextual replies very readily.

HAL will focus on the entire conversation to help it calculate what the next line of reply should be from those lines you have taught it (faster development will occur if you keep track of what those responses are and reuse them as much as possible). In contrast, its ability to parse syntax is currently limited to a single variable segment of any given sentence.

Teaching the HAL semantics is simply a matter of giving it enough varied conversations to allow it to build up a large enough set of data to allow it to predict where it is supposed to go next in its choice of dialogue, whereas teaching it syntax must primarily be done by rote; over time, slowly building up a matrix of similar sentences containing varied grammar segments.

  1  
'Send Send email to user    Reply with quote Reply with quote    Edit message Edit message

Forums Home    The Artificial Intelligence Forum    Alan and other chatbots    Language Mind and Consciousness  
Contact Us Terms of Use