The AI Engine

You are partially correct. You cannot create a bot, go away, and let it develop through talking to others. It is, however, different than ALICE in terms of offering more. If you look down in the bot contest forum, you will see an excellent post by Psimagus on the advantages of the PF (there are advanates in terms of seeks, plugins, spellcheck/grammar...etc.)

Maybe you also want to check out Ultra Hal (not on this site). That may be closer to what you are looking for.

thank you Bev . . I understand what you are saying and from what I have seen of the ALICE engine I think you are right . . the PF engine does seem to have a greater flexibility

I am a little familiar with Ultra Hal . . I had an early version of it

I have been looking at that . . and two other brain engines as possibilities to work with

the biggest problem I am having right now is finding a brain engine that will work with the quality of face/body/animation software that I would like to use

I have been tossing a chatbot idea around in my head for a while and now I just need to try and find the right combination of softwares to use together to try what I want to do

thank you very much for your feedback

I am getting xnones and no transcript.

I'm not getting xnones (touch wood) but I am not getting a transcript at all.

I found the transcrips..it's hooked to yesterdays?

"Not learning": Do you mean that they don't develop networks of associated information? I've had bots learn my interests and remember things I said. I assume the bot maker isn't hardwiring that from transcripts. They "learn" whether they like or dislike other people and bots. I haven't made it to the Advanced section of the Book yet, but I'm curious to know in what sense PF bots do not "learn."

They can learn just about anything if you have the memory for it..and someone will give them the info..names, sex, if they have a dog etc..The memory is where you can really make your bot shine..

Yes, memories are cool--also people who think about the logic how language is structured seem to make good bots (not me). The thing is, you still have to go over the transcripts so it's not like you set the bot up and just talk to it to make it learn (some other bots try to do that).

I like the PF a lot, so don't take that as a complaint. The PF has a different approach than some other "build a bot" places have. I think what the PF offers is better for what I want to do.

Devilferret-it would be great if the bots could do all those things. If you know someplace that would give the Prof a grant, or a company that would pay him money to add those tihngs, it would be great. I teased somebody about writing things like that a while back, so if you read that post, no we can't do that now.

BTW are you using People Putty or something else?

I'm curious to know in what sense PF bots do not "learn."

In a nutshell PF bots can learn facts (and even form opinions,) but not language use. Learning bots like Ally (written by our own Colonel720, and available at: http://www.be9.net/Ally/Index.html,) will actually learn their grammar and vocabulary from scratch, but require a LOT of patient training (during which time the level of conversation is at best curiously surreal, and at worst rather like trying to communicate with an autistic parrot.)

Quick question: if I want to have just an xnomatch seek for a keyphrase (ie, a seek that will always trigger), is there any way I can save the entire response that triggers it (like, as a memory)?

If, for example, the keyphrase was "anything * you like * here" I would imagine you could use something like:

{?PF remember "(prekey) anything (key1) you like (key2) here (postkey)" as "fullresponse"; ?}

MickMcA
Wed 10:12 AM April 12, 2006 Message: 5157




"Not learning": Do you mean that they don't develop networks of associated information? I've had bots learn my interests and remember things I said. I assume the bot maker isn't hardwiring that from transcripts. They "learn" whether they like or dislike other people and bots. I haven't made it to the Advanced section of the Book yet, but I'm curious to know in what sense PF bots do not "learn."


By not learning I meant that I did not want to have to create specific Q/A scripts in order for the bot to be able to respond to a stimulus.
For example . . if you give a PF bot a text document to read . . can it read and understand the document . . and then discuss that document with you?

To me . . if I have to write a series of Q/A scripts that the bot then uses to determine if it "likes" someone or not . . then the bot is not "learning" if it likes someone . . it is learning if "I" would like that particular text input.


Bev
Wed 12:56 PM April 12, 2006 Message: 5159

BTW are you using People Putty or something else?



I have looked at People Putty created faces and it reminds me of late 80's graphics . . low grade texturing and all.

What I want to try is Poser . . I have seen samples of what can be done in Poser and it is very nice quality.


psimagus
Wed 19 PM April 12, 2006 Message: 5160

I'm curious to know in what sense PF bots do not "learn."

In a nutshell PF bots can learn facts (and even form opinions,) but not language use. Learning bots like Ally (written by our own Colonel720, and available at: http://www.be9.net/Ally/Index.html,) will actually learn their grammar and vocabulary from scratch, but require a LOT of patient training (during which time the level of conversation is at best curiously surreal, and at worst rather like trying to communicate with an autistic parrot.)


To me . . for a bot to "learn" . . I should not have to go in and create a Q/A script for a specific type of expected data that will then store the input data into a memory array.
A bot that "learns" should be able to decide on it's own by understanding the input text string (or voice input) that I have just mentioned something that I "like" and then on it's own store that data.


What I am trying to avoid is creating a text scanning bot that will simply respond based on a pre-programmed script that the botmaster creates.